__Additions made to Stata during version 10__

This file records the additions and fixes made to Stata during the 10.0
and 10.1 releases:

+---------------------------------------------------------------+
| help file contents years |
|---------------------------------------------------------------|
| whatsnew Stata 15.0 and 15.1 2017 to present |
| whatsnew14to15 Stata 15.0 new release 2017 |
| whatsnew14 Stata 14.0, 14.1, and 14.2 2015 to 2017 |
| whatsnew13to14 Stata 14.0 new release 2015 |
| whatsnew13 Stata 13.0 and 13.1 2013 to 2015 |
| whatsnew12to13 Stata 13.0 new release 2013 |
| whatsnew12 Stata 12.0 and 12.1 2011 to 2013 |
| whatsnew11to12 Stata 12.0 new release 2011 |
| whatsnew11 Stata 11.0, 11.1, and 11.2 2009 to 2011 |
| whatsnew10to11 Stata 11.0 new release 2009 |
| **this file** Stata 10.0 and 10.1 2007 to 2009 |
| whatsnew9to10 Stata 10.0 new release 2007 |
| whatsnew9 Stata 9.0, 9.1, and 9.2 2005 to 2007 |
| whatsnew8to9 Stata 9.0 new release 2005 |
| whatsnew8 Stata 8.0, 8.1, and 8.2 2003 to 2005 |
| whatsnew7to8 Stata 8.0 new release 2003 |
| whatsnew7 Stata 7.0 2001 to 2002 |
| whatsnew6to7 Stata 7.0 new release 2000 |
| whatsnew6 Stata 6.0 1999 to 2000 |
+---------------------------------------------------------------+

Most recent changes are listed first.

-------- **more recent updates** --------------------------------------------------

See whatsnew10to11.

-------- **update 10jun2010** -----------------------------------------------------

__Stata executable, all platforms__

1. In rare circumstances, accessing the Internet from within Stata
could cause Stata to crash. This has been fixed.

__Stata executable, Mac (64-bit)__

2. Graphs exported from the 64-bit version of Stata for Mac to a bitmap
format, such as TIFF, would not be exported to a user-specified size
when Stata was run in Snow Leopard (Mac OS X 10.6). This has been
fixed.

__Stata executable, Ubuntu Linux__

3. In release 10.04 of the Ubuntu distribution of Linux, the data in
Stata's memory could become corrupt. This was caused by a change in
the behavior of a low-level call in an operating system library in
the Ubuntu distribution of Linux and so was restricted only to the
Ubuntu 10.04 distribution of Linux. This has been fixed.

-------- **update 20jan2010** -----------------------------------------------------

__Ado-files__

1. **svy: regress**, when fitting a model with duplicate *indepvars*,
reported the coefficient and standard error of the first element
occurring in **e(b)** for all of its duplicates in the coefficient
table. This misrepresented the actual fitted coefficients in **e(b)**
because typically all but one of the coefficients should have been
reported as "(dropped)". This has been fixed.

-------- **update 01oct2009** -----------------------------------------------------

__Ado-files__

1. **stci**, when used with prefix **by**, option **by()**, **if**, or **in**, produced
correct confidence intervals but reported standard errors that were
based on a survivor-function estimate that did not restrict the
sample. This has been fixed.

2. **sunflower** with option **addplot()** would only render the plots in
option **addplot()**. This has been fixed.

__Stata executable, all platforms__

3. Function **colnumb()** failed to match an equation specification on
columns containing time-series operators. This has been fixed.

4. Stata's old programmable dialog system was mistakenly disabled in
the 18aug2009 update. This functionality has been restored.

__Stata executable, Windows__

5. Applying the "Factory Settings" with some very specific windowing
arrangements could crash Stata. This has been fixed.

__Stata executable, 64-bit Mac__

6. The 18aug2009 update introduced a bug in the Graph Editor where
changing a selection in a pulldown menu of the graph toolbar would
have no effect on the graph being edited. This has been fixed.

-------- **update 04sep2009** -----------------------------------------------------

__Stata executable, Mac__

1. Changes in Mac OS X 10.6 (Snow Leopard) caused shell commands
executed from Stata to crash. Stata now accommodates these changes
in Snow Leopard when executing shell commands while maintaining
compatibility with earlier versions of Mac OS X.

__Stata executable, 64-bit Mac__

2. **sts graph** displayed a black background in the legend if the region
color was set to white. This has been fixed.

-------- **update 18aug2009** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 9(2).

2. Previously, **arch**, when used with option **het()**, did not display the
slope coefficients of the heteroskedasticity equation if the model
did not include any ARCH terms, though the coefficients were stored
in the coefficient vector **e(b)**. This has been fixed.

3. **areg**, in rare cases, would report a nonzero coefficient and a very
large standard error for a variable that had no within-group
variance instead of dropping the variable. This has been fixed.

4. **areg**, when used with models with very few observations per level of
the absorbed category, would occasionally exit with an error message
claiming insufficient observations even though the model was
estimable. This has been fixed.

5. **areg** reported an incorrect F test of the joint significance of the
absorbed indicators when a regressor was perfectly correlated with
them. This has been fixed.

6. **biprobit** with option **vce(robust)** or **vce(cluster ***clustvar***)** reported a
likelihood-ratio test for rho instead of a Wald test for rho. Also,
log likelihoods were reported in the iteration log instead of log
pseudolikelihoods. This has been fixed.

7. **blogit** and **bprobit**, when the number of positive responses was larger
than the total population, returned an error message. Observations
that have a number of positive responses that is larger than the
total population are now dropped from the analysis.

8. **egen** function **mode()** with option **missing** and a variable containing
all missing values incorrectly reported a warning message that
multiple modes were present. This has been fixed.

9. **estat classification**, when used after an estimation command with
**iweight**s, would treat the **iweight**s as **fweight**s if they were integer
valued or would report an error message if they were not integers.
**estat classification** now provides an appropriate error message any
time **iweight**s are used in the estimation.

10. **estat overid**, when used after **ivregress gmm**, would exit with an
error if the model included regressors with time-series operators.
This has been fixed.

11. **estat summarize** after **arch** and **arima** would appear to freeze Stata
and eventually exit with an error if the previous model did not
contain any ARCH or ARIMA terms. This has been fixed.

12. **glm** option **vce(unbiased)** could not be combined with **vce(cluster**
*clustvar***)** or with **vce(robust)**. They may now be combined (for
example, **vce(unbiased) vce(robust)**).

13. When option **vce(robust)** or **vce(cluster ***clustvar***)** was specified with
**heckprob** or the maximum-likelihood version of **heckman**, the test for
rho = 0 at the bottom of the output was labeled as a Wald test but
was actually a likelihood-ratio test. This has been fixed so that a
Wald test is really performed. When option **vce(robust)** or
**vce(cluster** *clustvar***)** was specified with these commands, a log
likelihood was reported in the iteration log. This has been changed
to a log pseudolikelihood.

14. **hetprob** with option **vce(robust)** or **vce(cluster ***clustvar***)** reported a
likelihood-ratio test for lnsigma2 = 0. Likelihood-ratio statistics
are not valid with robust VCEs. Therefore, this test is no longer
reported.

15. **ivregress**, when used with regressors that differed by many orders of
magnitude, could report erroneous results or drop regressors. Now
**ivregress** is more tolerant of poorly scaled data.

16. **lrtest**, when used after **xtmixed**, **xtmelogit**, or **xtmepoisson**, detects
differences in the number of estimated random-effects variances
between nested mixed models. Previously, if a difference was
detected, **lrtest** assumed that this was a result of testing the null
hypothesis that the variances in question were equal to zero, that
is, that the null hypothesis was on the boundary of the parameter
space. When this occurred, if the difference in model degrees of
freedom was equal to one, then the p-value was divided by two and
reported as chibar2(01). If the difference in model degrees of
freedom exceeded one, then a note was placed in the output stating
that the p-value was a conservative estimate. Such behavior was
correct most of the time, but not always.

Sometimes nested mixed models differ in the number of variances
estimated even when not testing on a boundary, for example, when
testing whether two variances are equal. In such situations, **lrtest**
either reported a chibar2(01) p-value that was anticonservative or
erroneously declared the p-value as a conservative estimate when, in
fact, the p-value was precise.

Rather than assume all tests are on the boundary or try to infer the
null hypothesis from the estimation results, **lrtest** now prints a
note stating when the possibility of a boundary test exists and
leaves it up to the user to determine whether this is the case. The
new behavior is guaranteed to be either conservative or precise, but
never anticonservative.

17. **nbreg** with option **vce(robust)** or **vce(cluster ***clustvar***)** reported a
likelihood-ratio test for alpha = 0. Likelihood-ratio statistics
are not valid with robust VCEs. Therefore, this test is no longer
reported.

18. The **nlogit** coefficient table did not display coefficients for
continuous variables associated with the base alternatives. This
has been fixed.

19. **predict, variance** after **arch** would sometimes exit with an error
message if the model was fit with constraints. This has been fixed.

20. **predict, xbd** and **predict, d** after **areg** changed the sort order of the
data. This has been fixed.

21. **predict** with option **stdp** would exit with an obscure error message
after **arima** if the model did not contain any regressors or a
constant term. Now a more descriptive error message is issued.

22. The predicted inclusive values, **predict, iv**, after fitting the RUM
consistent nested logit model, that is, **nlogit** without option
**nonnormalized**, were incorrect. This has been fixed.

23. **predict** after **streg** with option **frailty()** sometimes ignored options
**alpha1** and **unconditional** when used to predict quantities accumulated
over multiple-record data, such as cumulative martingale residuals
(option **cmgale**) and cumulative survival probabilities (option
**csurv**). This has been fixed.

24. **scobit** with option **vce(robust)** or **vce(cluster ***clustvar***)** reported a
likelihood-ratio test for alpha = 1. Likelihood-ratio statistics
are not valid with robust VCEs. Therefore, this test is no longer
reported.

25. **sts graph** would not honor suboption **cols()** of **legend()** when
specified with options **ci** and **by()**. This has been fixed.

26. **suest** with **mlogit** results that contained a period in an equation
name would report the "operator invalid" error. This has been
fixed. Now **suest** converts periods to commas.

27. **suest** with data that were **svyset** would report an error message if
any of the **svyset** variables was not found, even when option **svy** was
not specified and none of the estimation results was from prefix
**svy**. This has been fixed.

28. Undocumented command **svygen jackknife** produced an error message if
options **poststrata()** and **postweight()** were specified. This has been
fixed.

29. **treatreg**, when used with option **constraints()**, did not properly
apply the constraints when computing the likelihood-ratio test of
independent equations, resulting in an invalid test statistic. Now
**treatreg** reports a Wald test of independence when constraints are
applied.

30. **xthtaylor**, in rare cases, would declare a variable to be
time-invariant even though it was time varying. This has been
fixed.

31. **xtreg, fe** now reports a missing F statistic for the overall model
test when the VCE is less than full rank. The reported degrees of
freedom corresponds to the maximum number of coefficients that could
be simultaneously tested.

32. **xtreg, pa** with option **corr()** and **xtgee** with option **corr()**, when
there were more time periods than could be handled by **tabulate**,
returned an error message. This has been fixed.

33. **xtreg, re** and **xtgee** now report a missing Wald chi-squared statistic
for the overall model test when the VCE is less than full rank. The
reported degrees of freedom corresponds to the maximum number of
coefficients that could be simultaneously tested.

34. **xtreg, re** with option **vce(robust)** or **vce(cluster ***clustvar***)**, when the
dataset had balanced panels, reported an incorrect F statistic.
This has been fixed.

35. **zinb** with option **vce(robust)** or **vce(cluster ***clustvar***)** would report a
Vuong test or likelihood-ratio test of alpha = 0 if option **vuong** or
**zip** was also specified. Likelihood-ratio statistics and Vuong tests
are not valid with robust VCEs. Therefore, these tests are no
longer reported.

36. **zip** with option **vce(robust)** or **vce(cluster ***clustvar***)** would report a
Vuong test if option **vuong** was also specified. The Vuong test is
not valid with robust VCEs. Therefore, this test is no longer
reported.

37. **ztnb** with option **vce(robust)** or **vce(cluster ***clustvar***)** reported a
likelihood-ratio test for alpha = 0. Likelihood-ratio statistics
are not valid with robust VCEs. Therefore, this test is no longer
reported.

__Stata executable, all platforms__

38. Printing a selection that included the very last line of the Results
or Viewer windows caused Stata to crash. This has been fixed.

39. The Review window would show the incorrect return code for comments
that were typed interactively following an error message. This has
been fixed.

40. Mata nonuniform random-number generators, when views were used for
the parameter matrices, would crash Stata. This has been fixed.

41. Mata probability functions **binomialp()**, **hypergeometricp()**,
**nbinomialp()**, and **poissonp()**, when views were used for the parameter
matrices, would crash Stata. This has been fixed.

42. Mata functions **findexternal()**, **valofexternal()**, **crexternal()**, and
**rmexternal()** would allow names containing blanks if the name was
otherwise valid. This has been fixed.

43. Mata function **dir()** could crash Stata when the results were more
than 10,000 items. This has been fixed.

44. Mata probability distribution functions **hypergeometric()**,
**nbinomial()**, and **poisson()**, when views were used for the parameter
matrices, would crash Stata. This has been fixed.

45. Mata function **st_addvar()** would add a variable to the Stata dataset
even when the variable name contained blanks but was otherwise
valid. This has been fixed.

46. Mata function **st_isname()**, when the name being checked contained
blanks but was otherwise valid, returned valid (1). This has been
fixed.

47. Mata function **st_varrename()** would rename a variable to a name
containing blanks if the name was otherwise valid. This has been
fixed.

48. **svy****:** **mean**, **svy****:** **proportion**, **svy****:** **ratio**, and **svy****:** **total** would mark
out observations with missing values in the summary variables even
when the sampling weight was zero, which is a surrogate for
identifying out-of-subpopulation observations. This has been fixed.

__Stata executable, Windows__

49. **edit** or **browse** with an **if** expression that included all observations
in the dataset sometimes caused the vertical scrollbar to not
display or to scroll incorrectly. This has been fixed.

50. In the Data Editor, if the operating system failed to complete a
Clipboard copy, a memory leak could occur. This has been fixed.

51. **set linesize**, when the scrollbar automatically appeared in the
Results window, could be overwritten. This has been fixed.

52. **window manage associate** would fail to restore file associations
under Windows Vista if Stata was not started with administrator
access. This has been fixed.

53. In some circumstances, the Results window did not initialize the
linesize correctly when maximized inside Stata's main window. This
has been fixed.

54. The Review and Variables windows did not save and restore their font
settings with saved preferences. This has been fixed.

55. The Review window could show error codes next to the wrong command
if the Review window was initially closed when Stata was first
opened. This has been fixed.

56. "Compact Window Settings" did not autohide the Review and Variables
windows and did not correctly size Stata's main window. This has
been fixed.

57. In the Graph Editor, copying a graph to the Clipboard could cause
selecting a graph object to work incorrectly. This has been fixed.

__Stata executable, Mac__

58. In the Graph Editor, copying a graph to the Clipboard could cause
selecting a graph object to work incorrectly. This has been fixed.

59. After changing the aspect ratio of a graph from the Graph Editor,
the graph window would not automatically resize to the new aspect
ratio. This has been fixed.

__Stata executable, 64-bit Mac__

60. Depending on the number of lines displayed in the Results window,
**--more--** could still appear in the Results window even after the
more condition had been cleared. This has been fixed.

61. The **window fopen** and **window fsave** commands are now supported.

62. **File > Insert File...** will now insert files in the Do-file Editor.

63. Stata is now brought to the front when launched from the command
line.

64. If the Data Editor was started with an **if** condition and the Delete
dialog was opened more than once, Stata could crash. This has been
fixed.

65. How Stata redraws the graph window after the window has been resized
has been modified to accommodate changes to Mac OS X 10.6 (Snow
Leopard). This modification is also compatible with Mac OS X 10.5
(Leopard).

66. From the Graph Editor, adding a marker and then changing the
property of the newly added marker could cause Stata to crash. This
has been fixed.

67. The Graph Editor caused Stata to crash if the image lock16.png was
not found in Stata's ado-path or was corrupt. Although the
missing/corrupted image suggests that Stata's ado directory has been
corrupted and requires Stata to be reinstalled, Stata will now
ignore the missing icon to avoid crashing.

68. The current graph scheme was not being saved in Stata's preferences.
This has been fixed.

69. The Do-file Editor now automatically adds an end-of-line delimiter
at the end of a do-file if one does not already exist. Stata
requires that an end-of-line delimiter exist on all lines in a
do-file that are to be executed.

70. Stata did not convert filenames with Unicode characters from an Open
or Save dialog to a filename with the correct file system
representation. In some cases, this would result in "file not
found" errors. This has been fixed.

__Stata executable, Unix GUI__

71. **edit** or **browse** with an **if** expression that included all observations
in the dataset sometimes caused the vertical scrollbar to not
display or to scroll incorrectly. This has been fixed.

72. The Command window had a vertical scrollbar that was only displayed
when needed. Resizing the Command window so that it was shorter
than the height needed to display the entire vertical scrollbar
caused the Command window to grow in height to the minimum height
necessary to display the vertical scrollbar the next time Stata was
launched. This in turn also caused Stata's main window to grow.
Because of the resizing problem and the limited usefulness of the
vertical scrollbar, the Command window will now no longer display a
vertical scrollbar.

73. Selecting **File > Open** from a Do-file Editor window that had an open
do-file displayed the Open dialog with no current folder. This has
been fixed so that the Open dialog defaults to the directory of the
currently opened do-file.

-------- **update 16apr2009** -----------------------------------------------------

__Ado-files__

1. **estat concordance**, when used after **stcox** with the **strata()** option,
did not take into account the stratification structure of the data.
Thus, support for this feature has been temporarily disabled until a
statistically appropriate solution can be implemented.

2. **estat concordance** changed the sort order of the current dataset.
This has been fixed.

3. **fracpoly** and **mfp**, when used with **mlogit**, assigned incorrect model
degrees of freedom. The model degrees of freedom were too small,
resulting in smaller p-values. This has been fixed.

__Stata executable, Mac (32- and 64-bit)__

4. **edit** or **browse** with an **if** expression that included all observations
in the dataset sometimes caused the vertical scrollbar to not
display or scroll correctly. This has been fixed.

__Stata executable, Mac (32-bit)__

5. Graphs saved as a PDF sometimes had extra white space above the top
of the actual graph image in the PDF file. This has been fixed.

__Stata executable, Mac (64-bit)__

6. You can now sort the columns in the Review window. You can now drag
and drop selected commands from the Review window to a destination
that accepts text. You can also drag and drop selected variables
from the Variables window to a destination that accepts text.

7. Entering **doedit filename.do** from the command line when **filename.do**
does not exist creates a new Do-file Editor document and sets the
filename for the document. Although the document does not exist on
disk yet, selecting **File > Save...** for the first time for the new
document will save the document to disk without prompting for a
filename.

8. Programmable menu support (see **window menu**) is now available.

9. The graph **Save As** dialog now uses the current graph filename as the
default filename.

10. The plot type listbox would not show all plot types in some
situations. This has been fixed.

11. If Stata's main toolbar window was moved to an external display and
the external display was not available the next time Stata was
started, then the window did not appear on the primary display.
This has been fixed. Stata now allows the main toolbar window to be
dragged above the menu bar if an external display is arranged above
the primary display.

12. Opening the Data Editor with an **if** condition that returned no
observations (for example, **edit if foreign==.** when **foreign** had no
missing values) caused Stata to crash. This has been fixed.

13. The font preference for the Viewer window would not be permanently
saved until Stata had exited. This caused new Viewer windows during
the same session to not use the new font preference. This has been
fixed.

14. Stata did not adjust for the font leading when drawing text in
graphs. This would cause text to appear slightly lower than what
was intended. This has been fixed.

15. Stata would not save a TIFF graph. This has been fixed.

-------- **update 17mar2009** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 9(1).

2. **fracpoly** and **mfp** have been updated to support **stpm2** (Royston and
Lambert, forthcoming, *Stata* *Journal*), an improved version of
user-written command **stpm**.

3. **graph** with option **by(***varlist***)** failed if *varlist* was just one *varname*
and any of the levels of *varname* were either a blank string or an
empty label. This has been fixed.

4. **stsplit** would, on rare occasions, order survival times incorrectly
as the result of a precision bug. This occurred when your default
type was not **double** and you used **stsplit** to split records over
failure times that differed by less than float precision. This has
been fixed.

5. **sureg** with option **isure** could not replay its own estimation results.
This has been fixed.

-------- **update 06mar2009** -----------------------------------------------------

__Ado-files__

1. **suest** did not allow default options **robust** or **vce(robust)** to be
specified. This has been fixed.

__Stata executable, Mac (64-bit)__

2. The **Window** > **Do-file Editor** menu did not display a list of the
currently open Do-file Editor windows. This has been fixed.

3. The vertical scrollbar in the Do-file Editor would always move to
the top of the document after a new line was entered. This has been
fixed.

4. Using command **translate** to convert a SMCL file from disk to PDF
format would print the file instead. This has been fixed.

5. The Open button on the main toolbar would display only the most
recently opened files. It now displays the Open dialog when
clicked. Clicking and holding the left mouse button on the Open
button is required to display the Open Recents menu. The Save
button on the main toolbar now displays the Save As dialog when
there are data in memory.

6. When running or doing an unsaved do-file from the Do-file Editor and
autosave is enabled, you may be presented with the warning "This
document's file has been changed by another application since you
opened or saved it" when the do-file is saved or run. This has been
fixed and you should no longer see the warning.

7. Some lines in graphs would not be drawn with the correct thickness.
This has been fixed.

-------- **update 26feb2009** -----------------------------------------------------

__Ado-files__

1. **alpha**, an r-class command, cleared all results stored in **e()**. This
has been fixed.

2. **alpha** returned a "no observations" error when the sense could not be
determined empirically. An error is now given suggesting that
option **asis** be specified.

3. **test** after **anova** did not allow options **test()** and **matvlc()** to be
specified together. This has been fixed.

__Stata executable, Mac (32-bit)__

4. In the previous update, Stata changed its behavior from copying
graphs to the Clipboard in either the PDF or PICT format to copying
graphs in multiple graphic formats, including PDF, TIFF, and PICT.
This allows both modern and legacy applications to paste graphs from
Stata in a format the applications prefer. However, Microsoft
Office 2004 does not support multiple graphic formats in the
Clipboard. Stata has reintroduced the "Copy images to the Clipboard
only as PICT" option in the General Preferences dialog to copy
graphs to the Clipboard only in the PICT format.

5. The Variables window scrollbar no longer moves to the beginning of
the variables list when variables are added.

__Stata executable, Mac (64-bit)__

6. You can now drag and drop Stata graphs to the Desktop and to
applications that support the PDF or TIFF format.

7. Clicking on a variable name or an observation number in the Data
Editor did not select a variable's column or observation's row,
respectively. This has been fixed.

8. Double-clicking on a Stata file did not automatically change the
current working directory within Stata. This has been fixed.

9. Stata required a restart after its Internet proxy settings were
changed before they would take effect. Stata's Internet proxy
settings now take effect immediately after they have been changed.

10. The Viewer would not use the color scheme preference from a previous
session. This has been fixed.

-------- **update 02feb2009** -----------------------------------------------------

__Ado-files__

1. The dialog box for creating display formats contained a sample for
weekly formats that was invalid. This has been fixed.

2. **mkspline**, when used to produce linear splines with equally spaced
knots, could create spline variables containing all 0s if the knots
could not be uniquely identified when formatted as **%9.0g**. This has
been fixed.

3. **sktest** previously used casewise deletion. It now uses variablewise
deletion.

4. The dialog box for **xtdpdsys** did not allow options **twostep** and
**vce(robust)** to be specified together. This has been fixed.

__Stata executable, all platforms__

5. **describe**'s **varlist** option, when the data were not sorted, reported
the entire varlist, or the given varlist if specified, in
**r(sortlist)**. This has been fixed.

6. **estimates use** misread **logistic**, **logit**, and **probit** estimation results
that identified multiple perfect predictors, variables dropped
because of collinearity, or both, possibly resulting in Stata
crashing. This has been fixed.

7. The estimation algorithm for **exlogistic** has been improved to use
relative frequencies. This means that it can successfully fit
models with larger datasets and higher frequencies of enumerated
sufficient statistics. The error message has also been improved
when the datasets or sufficient statistics are too large.

8. **exlogistic** with one or more independent variables collinear with the
dependent variable terminated with an error message. A check for
collinearity is unnecessary and has been removed.

9. Mata's **mata rename** did not check whether the new matrix name was
valid. This has been fixed.

10. Mata's **st_local()** function, after reading a string from a binary
file, would sometimes return a string with garbage values appended
to the end of the original string. This has been fixed.

11. **matrix accum** without option **noconstant** and with one less variable
than the current value of **c(matsize)** caused Stata to crash. This
has been fixed.

12. **odbc insert** with option **create** would issue an error message when
inserting data stored as a double into Excel. This has been fixed.

13. **odbc query** now has two new options, **verbose** and **schema**. **verbose**
lists any data source alias, nickname, typed table, typed view, and
view along with tables so that you can load data from these table
types. **schema** lists schema names with the table names if the data
source returns schema information.

14. **predict** with **matsize** set smaller than required for the current
estimation results caused Stata to crash. This has been fixed.

15. **syntax** incorrectly parsed wildcard (*****) options when arguments
contained unbalanced parentheses enclosed in double quotes; **syntax**
parsed them correctly when enclosed in compound double quotes. It
now works correctly in both cases.

__Stata executable, Windows__

16. The Do-file Editor, while the Data Editor was open, did not redraw
correctly. This has been fixed.

17. Printing a do-file by selecting **File** > **Print** from the main Stata
window crashed Stata. This has been fixed.

18. The drop-down list in the Graph window's "Save as" dialog box did
not correctly change the file extension, for example, from ***.gph** to
***.emf**. This has been fixed.

19. On the Graph window, the toolbar did not remember customizations
between sessions. This has been fixed.

__Stata executable, Mac__

20. Stata now includes both PDF data and PICT data when copying Stata
graphs or Stata output to the Clipboard. This will allow
applications that do not support PDF data to still paste data from
the Clipboard in the PICT format. As a consequence of this change,
the "Copy images to the Clipboard as PICT" setting is no longer
necessary and has been removed.

21. Stata now uses the default end-of-line delimiter preference when
copying text to the Clipboard from the Results, Viewer, and Data
Editor windows. This is necessary for applications such as
OpenOffice that only support the Unix end-of-line delimiter in text
pasted from the Clipboard.

22. Stata previously used the incorrect uniform type identifier (UTI)
description of "ASP Code file" for the different text files it
supports such as **.do** and **.ado**. The mislabeled description did not
affect the functionality of the text files. Stata now uses a unique
UTI for each text file it supports, for example, "Stata Do-file" for
do-files and "Stata Ado-file" for ado-files. This is useful for
distinguishing between Stata's different text files in applications
such as a Spotlight search. The additional UTIs may require
rebuilding the LaunchServices database on your Mac before the
changes become visible, but it is not urgent. Please contact
technical support if you require further assistance in rebuilding
the database.

23. The Variable window's scrollbar would not reset its position when
data were cleared or when new data were used. This has been fixed.

24. Stata would not open a file in a web browser when using the SMCL
directive **browse**. This has been fixed.

__Stata executable, Unix__

25. For some Unix platforms, such as the Solaris Sparc, printing the
Results window by using the "Print" dialog box caused Stata to
crash. This has been fixed.

-------- **update 06jan2009** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 8(4).

2. Stata's **%tC** date-and-time format now includes a leap second
following 31dec2008 23:59:59, namely, 31dec2008 23:59:60, as
recently specified by the International Earth Rotation and Reference
Systems Service (IERS). The update will not take effect until Stata
is restarted after the update is installed.

3. **collapse** with small **aweight**s occasionally produced incorrect
percentile values. This has been fixed.

4. **egen** function **group()** with the **label()** option failed if a value
label contained double quotes. This has been fixed.

5. **estat bootstrap**, when used with certain estimation commands and when
the expression list was something other than **_b**, sometimes
incorrectly reported "estimation results not found". This has been
fixed.

6. **estat gof**, when used after a **probit**, **logit**, or **logistic** command that
included either the **vce(robust)** or the **vce(jackknife)** option,
reported an incorrect test result. The test result was incorrect
because it used the linear predictor in place of the predicted
probability of success. This has been fixed.

7. In all graphs, value labels and string variables are handled more
elegantly. Specifically, labels and string variables that contain
multiple quoted strings, such as

. **label define mylbl 1 `""this" "and that""'**

now follow the standard graphics convention and produce multiline
subtitles on by graphs and produce multiline axis tick labels when
the **valuelabel** option is specified.

This change also allows labels or strings to include unbalanced
parentheses and leading equal signs, which previously could cause
the **graph** command to fail.

8. **hausman**, following estimation with **vce(robust)**, sometimes reported a
test statistic and significance level, even though it was not
statistically appropriate to do so. **hausman** now issues an error
message instead.

9. **ivtobit** with the two-step estimator and multiple endogenous
regressors reported incorrect chi-squared statistics for the test of
exogeneity. This has been fixed.

10. **ksmirnov** with the **exact** option reported negative p-values on rare
occasions because of a precision problem. This has been fixed.

11. **linktest**, after estimation with the **svy** prefix, produced standard
errors that did not take into account the survey design. **linktest**
is no longer available after the **svy** prefix.

12. **predict**, after **streg** with the **svy** prefix, no longer allows the
**csnell**, **mgale**, **deviance**, **ccsnell**, and **cmgale** options. These
diagnostic predictions are statistically inappropriate for survey
data.

13. **rolling** no longer limits the number of panels allowed based on the
number of characters allowed in a macro.

14. **rolling**, when receiving an error from a command for a certain subset
of data, would post missing values in two observations instead of in
one. This has been fixed.

15. **ssc type** now allows new option **asis** to display files without
interpretation of SMCL directives.

16. **stcox**, when used with the **svy** prefix, no longer allows the
diagnostic options **mgale()**, **schoenfeld()**, and **scaledsch()**. These
diagnostics are statistically inappropriate for survey data.

17. **sts list** with the **saving()** option did not allow spaces in filenames.
This has been fixed.

18. **table**, when **aweight**s were specified and when missing values were
present, incorrectly computed the **semean** statistic. This has been
fixed.

19. **xtdescribe** treated missing values in the panel ID variable as a
valid group rather than marking the observations out of the sample.
This has been fixed.

20. **xtile** with ties occasionally produced incorrect quantile categories.
This has been fixed.

-------- **update 15oct2008** -----------------------------------------------------

__Ado-files__

1. **estat correlation**, after **asmprobit** and **asroprobit**, now uses a
default output format of **%9.4f** instead of the previous **%6.3f**.

2. In **estimates table** the **stfmt(%***fmt***)** option no longer affects the
display format of the estimation sample size, N.

3. **icd9**'s database has been updated to use codes through the V26 update
of 1 October, 2008.

4. **ssc hot** now provides more informative error messages if certain
files stored at the SSC are not as expected, although you will never
see these messages. These improvements were requested by Kit Baum,
who maintains the SSC site.

5. **tab2** has a new option, **firstonly**, that restricts the output to only
those tables that include the first specified variable. Use this
option to obtain tables of one variable interacted with a set of
other variables.

6. **tsappend** with panel data is now faster.

7. **xtmelogit** and **xtmepoisson** did not save **e(converged)** as documented.
This has been fixed.

-------- **update 22sep2008** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 8(3).

2. **codebook** with **if** or **in** qualifiers and option **compact** presented the
wrong number of observations for string variables. This has been
fixed.

3. The **egen** function **total()** with the **missing** option gave results that
depended on the sort order. This has been fixed.

4. **heckman** with option **twostep**, when regressors appeared in both the
outcome and selection equations, reported a Wald chi-squared test
for the coefficients in both equations. This has been fixed so that
now a joint test is performed for only those coefficients in the
outcome equation.

5. **heckman** with option **first**, when no dependent variable was specified
in the selection equation, displayed a temporary variable name in
the output. This has been fixed.

6. **ksmirnov** now issues the appropriate error message when used with a
string variable.

7. **prtesti** with option **count** did not allow specifying equal counts
(equivalent to specifying a proportion of 1), contrary to the
documentation. This has been fixed.

8. **spearman** with option **pw** produced incorrect values of correlation
coefficients and significance levels for pairs of variables
containing missing values when more than two variables were
specified. This has been fixed.

9. **stphplot** with option **adjust()** produced an error message when
**pweight**s were specified with **stset**. This has been fixed.

10. **sts list** option **saving()** can now be combined with options **failure**
and **at()**.

11. **sts list** with option **saving()** produced an error message when the
**compare** option was specified to compare more than six groups. This
has been fixed.

12. **svy** with an **if** expression in the **subpop()** option could result in the
error "no observations in subpop() subpopulation" as a result of the
update on 11aug2008. This has been fixed.

13. **xtreg** now allows the **in** syntax for specifying a range of
observations.

14. **xtreg** did not accept the** inrange()** function as part of an **if**
expression. This has been fixed.

15. **xtreg**, when the panel variable was included in the model as an
independent variable, sometimes returned a confusing message. This
has been fixed.

16. **xtreg** did not check for collinearity between the independent
variables and the panel variable. This has been fixed.

-------- **update 26aug2008** -----------------------------------------------------

__Ado-files__

1. **reshape**, when reshaping from long to wide format with xij variable
names longer than 20 characters, produced an error message and
stopped. This has been fixed.

2. **svy**'s descriptive statistics commands (**svy: mean**, **svy: proportion**,
**svy: ratio**, and **svy: total**) with the **over()** option would exit with
the error message "conformability error" when one or more over
groups were dropped because of missing value patterns in the
varlist. This has been fixed.

3. **svy**'s descriptive statistics commands (**svy: mean**, **svy: proportion**,
**svy: ratio**, and **svy: total**) were marking out observations that had
missing values in the **over()** variables for observations outside the
subpopulation. This affected the estimated variance values when the
missing value patterns resulted in dropped primary sampling units,
decreasing the design degrees of freedom. Both of these effects
were very slight and inversely related to the number of PSUs. This
has been fixed.

4. **svy: tabulate** was marking out observations that had missing values
in the variable list for observations outside the subpopulation.
This affected the estimated variance values when the missing value
patterns resulted in dropped primary sampling units, decreasing the
design degrees of freedom. Both of these effects were very slight
and inversely related to the number of PSUs. This has been fixed.

-------- **update 11aug2008** -----------------------------------------------------

__What's new in release 10.1 (compared with release 10.0)__

Highlights for release 10.1 include the following:

**o** Distribution and probability mass functions for the
hypergeometric, negative binomial, and Poisson distributions have
been added to Stata. The probability mass function for the
binomial distribution has also been added. See density functions.
Additionally, Mata now has these functions; see **mata normal()**.

**o** Random-variate functions for the beta, binomial, chi-squared,
gamma, hypergeometric, negative binomial, normal, Poisson, and
Student's t distributions have been added to Stata; see
random-number functions. Additionally, Mata now has these
random-variate functions; see **mata runiform()**.

**uniform()** has been renamed **runiform()** to be consistent with the
naming convention of the nonuniform random-variate functions.
**uniform()** continues to work but is now undocumented. Mata
functions **uniform()** and **uniformseed()** have been renamed **runiform()**
and **rseed()**.

**o** **asmprobit** and **asroprobit** have several new features, including
option **factor(***#***)** specifying the use of a factor covariance
structure, which models the covariance as I + C'C. Postestimation
command **estat facweights** displays the covariance factor weights in
matrix form.

**o** **reshape** now preserves the J variable value and variable labels and
all xij variable labels when reshaping from wide to long and back
to wide.

**o** **pwcorr** has new option **listwise**, which specifies that missing
values are to be handled by listwise deletion, allowing users of
**pwcorr** to mimic **correlate**'s treatment of missing values while
maintaining access to all of **pwcorr**'s features.

**o** Initialization Mata functions **ghk_init()** and **ghkfast_init()**, as
well as several other helper functions, have been added to
construct, modify, and query structures defining simulation
parameters and creating simulation points for subsequent calls to
**ghk()** and **ghkfast()**.

We recommend that you set the version to **version 10.1** at the top of new
do-files and ado-files. See **version**.

__Ado-files__

1. **asmprobit** and **asroprobit** have a new covariance structure. Option
**factor(***#***)** specifies the use of factor covariance structure. For a
model with J alternatives, the *#* x J factor matrix C, *#* smaller than
J, models the covariance as I + C'C, where I is a J x J identity
matrix.

2. **asmprobit** and **asroprobit** have new option **favor(speed**|**space)**, which
directs the programs to favor either speed or space (memory) when
executing. **favor(speed)** is the default.

3. **asmprobit** and **asroprobit** have new option **nopivot**, which will direct
the programs not to pivot integration intervals so that wider
intervals are on the inside of the multivariate integration. This
option may be useful when fitting a model with few simulation
points.

4. **biplot**, when computing means (and standard deviations with option
**std**) used in the computations, did not honor **if** or **in** qualifiers.
This has been fixed.

5. **bootstrap** and **jackknife** removed option **vce()** when it was specified
on the command being bootstrapped or jackknifed, resulting in
calculations based on the conventional standard errors and
covariances instead of the requested alternative. This has been
fixed.

6. **canon**'s default output has changed. **canon** now displays matrices of
raw coefficients by default. **canon** no longer documents option
**coefmatrix**, which specified the display of matrices of raw
coefficients. Setting the version to less than 10.1 will display
the raw coefficients, conditionally estimated standard errors, and
confidence intervals in the standard estimation table, by default;
this output can also be obtained by using new option **stderr**.

7. **dstdize** reported large population totals rounded to floating point
precision. This has been fixed.

8. **egen** function **mode()** with option **missing** now treats missing values
as a category when finding the mode. The **minmode**, **maxmode**, and
**nummode()** option scan now be specified with and without the **missing**
option for the desired mode. When **version** is set to less than 10.1
with option **missing**, missing values are not treated as a category.

9. **egen** functions **total()** and **rowtotal()** have new option **missing**. With
this option, if all values of *exp* or *varlist* for an observation are
missing, then that observation in *newvar* will be set to missing.

10. New postestimation command **estat facweights** after **asmprobit****,**
**factor(***#***)** and **asroprobit****, factor(***#***)** displays the covariance factor
weights in matrix form.

11. The dialog box for **fdasave** did not place quotation marks around the
path and filename, causing an error when the path or filename
contained spaces. This has been fixed.

12. In the Graph Editor, when editing a bar graph created with the *by*
*option*, the Sort button of both the Contextual Toolbar for the bar
region and the properties dialog for the bar region failed to sort
the graph's bars. This has been fixed.

13. The dialog box for **graph twoway** now allows plots to be reordered
when multiple plots have been defined.

14. **ivregress**, when a heteroskedasticity-and-autocorrelation consistent
VCE or weight matrix was requested with the Bartlett kernel and
Newey and West's optimal bandwidth-selection algorithm, could report
incorrect standard errors. When the GMM estimator was used, the
point estimates could also be incorrect. This has been fixed.
Whether you observe changes in output as a result of this bug fix is
data dependent.

15. Mata functions **ghk()** and **ghkfast()** have a new syntax.
Initialization functions **ghk_init()** and **ghkfast_init()**, as well as
several helper functions, have been added to construct, modify, and
query structures defining simulation parameters and creating
simulation points for subsequent calls to **ghk()** and **ghkfast()**.

16. **pwcorr** has new option **listwise**, which specifies that missing values
are to be handled by listwise deletion; that is, observations
containing missing values for any of the given variables are removed
from the estimation sample before any calculations take place. By
default, **pwcorr** uses pairwise deletion, under which the correlation
calculation for each variable pair uses all available data without
regard to missing values outside that variable pair.

Specifying **listwise** allows users of **pwcorr** to mimic **correlate**'s
treatment of missing values while maintaining access to all of
**pwcorr**'s features.

17. **reshape** now preserves the J variable value and variable labels and
all xij variable labels when reshaping from wide to long and back to
wide. Previous behavior is obtained by setting **version** to less than
10.1.

18. **roctab** and **roccomp**, although r-class commands, left information in
the **ereturn list** from calls they made to the **logistic** command. This
has been fixed.

19. **stpower** did not allow spaces in filenames specified in option
**saving()**. This has been fixed.

20. **svy**'s descriptive statistics commands (**svy: mean**, **svy: proportion**,
**svy: ratio**, and **svy: total**) were marking out observations that had
missing values in the variable list for observations outside the
subpopulation. This affects the estimated variance values when the
primary sampling units were the individual observations and could
decrease the design degrees of freedom. Both of these effects are
very slight and inversely related to the sample size. This has been
fixed.

21. **svy**, in the **if** expression within the **subpop()** option, did not allow
the **missing()** function to be specified. This has been fixed.

22. **xthtaylor**, when used with a sufficient number of perfectly collinear
covariates, would drop different variables when the same command was
repeatedly run. This has been fixed.

23. **xtgee**, when used with **corr(fixed** *matname***)**, deleted matrix *matname*.
This has been fixed.

24. **xtmelogit** and **xtmepoisson** now support maximize option **from()**; see
maximize. This allows users to specify starting values, which can
significantly speed up estimation. For these two commands, **from()**
is particularly useful when combined with **refineopts(iterate(0))**,
which bypasses the initial optimization stage.

25. As of Stata 10.1, **xtmixed**, **xtmelogit**, and **xtmepoisson** now require
that random-effects specifications contain an explicit level
variable (or **_all**) followed by a colon. Previously, if these were
omitted, a level specification of **_all:** was assumed, leading to
confusion when only the colon was omitted. To avoid this confusion,
omitting the colon now produces an error, with previous behavior
preserved under **version** control.

26. **xtreg** with option **be**, when used with regressors that included
time-series operators, would exit with an error message saying
"time-series operators not allowed". Now time-series operators are
allowed.

__Stata executable, all platforms__

27. Distribution and probability mass functions for the hypergeometric,
negative binomial, and Poisson distributions have been added to
Stata. The probability mass function for the binomial distribution
has also been added. See density functions. Additionally, Mata now
has these functions; see **mata normal()**.

28. Random-variate functions for the beta, binomial, chi-squared, gamma,
hypergeometric, negative binomial, normal, Poisson, and Student's t
distributions have been added to Stata; see random-number functions.
Additionally, Mata now has these random-variate functions; see **mata**
**runiform()**.

29. Function **uniform()** has been renamed **runiform()** to be consistent with
the naming convention of the nonuniform random-variate functions.
**uniform()** continues to work but is now undocumented. Mata functions
**uniform()** and **uniformseed()** have been renamed **runiform()** and
**rseed()**.

30. Old function **Binomial()** is no longer allowed (as of Stata 10.1);
**binomialtail()** should be used instead. **Binomial()** is allowed when
**version** is set to less than 10.1.

31. **drawnorm** now uses new function **rnormal()** to generate normal random
variates. When **version** is set to less than 10.1, **drawnorm** reverts
to using **invnormal(uniform())**.

32. Mata's addition colon operator, **:+**, crashed Stata if one of the
arguments was a void string matrix. This has been fixed.

33. Mata function **hash1(x, n)** incorrectly returned n+1 for certain x and
n. This has been fixed.

34. Mata function **strofreal()** returned different results depending on
the style of decimal point setting. This has been fixed.

35. **mlmatbysum** now checks for string variables in **by()**. String
variables are not allowed in this option.

36. **odbc insert** with option **create** now writes Stata doubles as SQL data
type FLOAT(53) instead of REAL. Most of the time in SQL, the REAL
data type maps to a 4-byte float, so on SQL Server, for example,
there was a loss of precision for all doubles.

37. **oneway**, with a constant *response_var* and option **bonferroni**, **scheffe**,
or **sidak**, produced invalid p-values in the comparison table. For
larger datasets, Stata stopped responding while **oneway** was in the
middle of producing the comparison table. This has been fixed.

38. **stcox** now allows a maximum of 100 variables to be specified in
option **tvc()**. The previous limit was 10.

__Stata executable, Windows__

39. **postfile** with option **every()** did not cause **post** to write results to
disk as often as was specified in **every()**. This has been fixed.

40. The Do-file Editor did not allow printing a selection. This has been
fixed.

41. While grid editing in the Graph Editor, dragging an object into
another cell did not highlight the target cell properly. Now the
target cell is shown in a darker translucent red, which is
consistent with Stata for Mac OS and Unix.

__Stata executable, Mac__

42. You can now disable keyboard navigation in the Variables and Review
windows from the General Preferences dialog. See revkeyboard or
varkeyboard for more information.

43. Pressing Cmd+Left arrow and Cmd+right arrow in the Command window
will now move the cursor to the beginning of a line and to the end
of a line, respectively.

__Stata executable, Unix__

44. **printer define** now allows up to 100 defined printers. The previous
limit was 10.

45. Previously, when opening an existing do-file using File > Open from
an existing Do-file Editor window, the do-file to be opened would
replace the contents of the existing Do-file Editor window,
regardless of whether the existing window had been saved. Stata now
always opens an existing do-file in a new window. The Open Recent
menu would display some characters as underlined if a filename
contained an underscore. This has been fixed. The Open Recent menu
in a new Do-file Editor window would not show the most recent file
added to the menu. This has been fixed.

-------- **update 25jun2008** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 8(2).

2. **collapse** and **table** now allow the statistics **semean**, **sebinomial**, and
**sepoisson** to calculate the standard error of the mean, the binomial
standard error of the mean, and the Poisson standard error of the
mean, respectively.

3. **collapse** with the *target_var***=***varname* syntax and with an empty weight
(that is, **[]**) produced an error message. This has been fixed.

4. **collapse** with** aweight**s and when ties were present on rare occasions
produced inconsistent percentile results. This has been fixed.

5. **median** did not allow a string variable in option **by()**. This has
been fixed.

6. **ml**'s documentation mistakenly contained **bootstrap** and **jackknife**
among the *vcetype*s allowed in the **vce()** option. This has been fixed
in the online help.

7. **nestreg** would fail to recognize the correct estimation sample for
**stcox** and **streg** when **stset** marked out observations, such as for
missing values in *timevar* or when the *if* qualifier was specified.
In such cases, **streg** would report the error "# obs. dropped because
of estimability, this violates the nested model assumption" and
quit. This has been fixed.

8. **svy brr** and **svy jackknife**, when used to prefix r-class commands,
would exit with the error "statistic ... evaluated to missing in
full sample". This has been fixed.

9. **svy jackknife**, when a string variable was specified in the **strata()**
option of **svyset** (without the **jkrweight()** option), reported a "type
mismatch" error. This has been fixed.

10. **table** with option **contents(freq)** did not allow more than seven
digits in the **format()** option. This has been fixed.

-------- **update 30may2008** -----------------------------------------------------

__Ado-files__

1. **dfuller** with option **regress** overwrote the estimation results with
those from the Dickey-Fuller regression. **dfuller** now leaves
existing estimation results unchanged.

2. **estat endogenous** and **estat overid** now save in **r()** the p-values from
chi-squared and F tests.

3. **estat firststage** after **ivregress** exited with an error when the model
contained time-series operators. This has been fixed.

4. **ivregress 2sls** and **ivregress gmm** have new option **perfect**, which
skips checking whether the endogenous regressors are collinear with
the excluded instruments.

5. **tsset** and **xtset** ignored period options such as **daily** or **weekly** if
the time variable had previously been formatted with a different
period. These commands now honor the specified period option and
issue a warning if it requires resetting the period.

6. **xtlogit** with option **fe** ignored option **from()**. This has been fixed.

-------- **update 27may2008** -----------------------------------------------------

__Ado-files__

1. The **bootstrap****:** and **jackknife****:** prefixes now allow you to specify the
**vce()** option on the command being bootstrapped or jackknifed. This
allows you to bootstrap or jackknife alternate estimates of standard
errors and covariances.

2. The dialog box for **describe** could sometimes initialize with
checkboxes in the incorrect state. This has been fixed.

3. **estat anova** after **discrim lda** did not abbreviate long variable names
causing misaligned output. This has been fixed.

4. **exlogistic** and **expoisson** now issue an error message when the
requested memory (option **memory()**) exceeds the limit of two
gigabytes.

5. **exlogistic** and **expoisson** in rare cases declared nonconvergence on
calculations for confidence bounds when in fact finite bounds were
computable. This has been fixed.

6. Mata's **optimize()** with *singularHmethod* **"hybrid"** did not always
optimally recover from bad steps. This has been fixed.

7. **mfp**, when used to fit an **xtgee** model with a single covariate,
reported a missing deviance. This has been fixed.

8. **ml display** has new option **nocnsreport**, which suppresses the display
of constraints above the coefficient table.

9. **nlcom** and **testnl** no longer require that **e(cmd)** exists. This allows
both to work with a wider range of user-written estimation commands.

10. **statsby** with option **subsets** and a string **by()** variable produced a
type mismatch error. This has been fixed.

11. **sts graph** with options **tmin()** or **tmax()** and option **risktable** could
produce incorrect at-risk counts within the table. This has been
fixed.

12. **sts list** has new option **saving()**, which will create a dataset
containing the results.

13. **sts list** does not support standard errors and confidence intervals
when the data are **stset** with **pweight**s. However, when option **at()**
was also specified, standard errors and confidence intervals were
reported, but these were inappropriate for sampling-weighted data.
As such, they are no longer reported.

14. **xtcloglog**, **xtfrontier**, **xtintreg**, **xtlogit**, **xtnbreg**, **xtpoisson**, and
**xtprobit** now issue more informative error messages when invalid
**vce()** types are specified.

15. **xtmelogit** and **xtmepoisson**, when used to fit a crossed-effects model
that includes both factor notation (**R.***varname*) and random
coefficients, exited with a matrix conformability error. This has
been fixed.

16. **xtpoisson** with option **fe**, when used with covariates that are
collinear with the panel identifier, incorrectly labeled
coefficients on variables that followed the dropped collinear
variables. This has been fixed.

17. **xttab** now returns the matrix of results in **r(results)** and the number
of panels in the sample in **r(n)**.

18. **xttab**, when used on data with unbalanced panels, produced
within-panel tabulations that incorrectly treated the panels as
balanced. This has been fixed.

__Stata executable, all platforms__

19. **c(smallestdouble)** has been added to **creturn list**.

20. Mata has the following new functions.

A. **asarray()** provides associative arrays, also known as containers,
maps, dictionaries, indices, and hash tables; see **[M-5]**
**asarray()**.

B. **[M-5] Dmatrix()** creates duplication matrices; see **[M-5]**
**Dmatrix()**.

C. **hash1()** implements Jenkins' one-at-a-time hash function; see
**[M-5] hash1()**.

D. **[M-5] Kmatrix()** creates commutation matrices; see **[M-5]**
**Kmatrix()**.

E. **[M-5] Lmatrix()** creates elimination matrices; see **[M-5]**
**Lmatrix()**.

21. Mata functions **[M-5] vec()** and **[M-5] vech()** have been sped up for
real and complex matrices.

22. Mata's **smallestdouble()** value was twice as large as the smallest
full-precision double. This has been fixed.

23. Stata failed to recognize matrix subscripts when they were strings
with leading spaces. This has been fixed.

24. **odbc load**, when converting dates, failed to fully honor version
control. When the version was set to 9 or less, it converted dates
to the newer **%tc** format rather than the older **%td** format. This has
been fixed.

25. Function **regexr()**, when called with an empty string as its first
argument, would sometimes incorrectly return the result of its
previous call. This has been fixed.

26. **smallestdouble()** has been added to Stata's programming functions.
It returns the smallest full-precision double.

27. **xtabond**, **xtdpd**, and **xtdpdsys**, when used with data that contained
gaps in panels such that the gaps varied from panel to panel,
produced estimates that were based on an incorrect instrument
matrix. This has been fixed.

28. The Data Editor, when used with many variables containing long
strings, would occasionally draw incorrectly and occasionally have
difficulty scrolling horizontally. These problems have been fixed.

29. The Data Editor now respects display formats on string variables.
That is to say, what you see in a cell is the formatted string.

30. When Stata closed, it did not call the class destructors for class
instances that existed at exit. This affected a few sophisticated
programmers who needed to know when every instance was destroyed so
that they could perform explicit clean-up operations for their
instances. This has been fixed.

__Stata executable, Windows__

31. Executables for Stata and StataAdministrativeTools are now signed,
allowing Windows Vista to identify StataCorp LP as the software
publisher. Previously, starting Stata with Administrator access and
with User Access Control (UAC) enabled would cause a warning to be
displayed that identified Stata as an "Unidentified Publisher".

32. The extended **macro** function **dir** has the new option **respectcase**,
which specifies that **dir** respect the case of filenames when
performing matches. Windows has case-insensitive filenames. As
such, by default under Windows **dir** converts all filenames to
lowercase before matching them. **respectcase** prevents this lowercase
conversion.

33. **odbc** in 64-bit Stata for Windows would crash when loading the
Windows 64-bit ODBC driver. This has been fixed.

34. The Do-file Editor, when opening a file, would place the cursor at
the end of the file instead of at the beginning. This has been
fixed.

35. The Do-file Editor did not open the file menu when Alt+F was
pressed. This has been fixed.

36. The Graph Editor did not always populate the Object Browser when the
Browser was opened manually. This has been fixed.

__Stata executable, Mac__

37. **--more--** conditions in the Results window unnecessarily caused high
CPU usage until cleared. This has been fixed.

38. Selecting a file using **window fopen** could cause Stata to crash if
the file's path was greater than 32 characters. This has been
fixed.

__Stata executable, Unix__

39. **log** with option **append**, when used to create a new file rather than
appending to an existing file, would apply incorrect permissions to
the newly-created file. This has been fixed.

-------- **update 02apr2008** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 8(1).

2. **nestreg** reported the error "arguments to title do not match the
number of columns" while displaying its summary of results when used
with **regress** with the **vce(bootstrap)** option. This has been fixed.

3. **suest** complained about missing score values with results from
**heckman** or **heckprob** when the independent variables in the main
equation (but not in the selection equation) contained missing
values outside the selection observations. This has been fixed.

4. **[SVY] svy** now mentions **e(b)** and **e(V)** in the **Stored results** section.

5. svy's linearized variance estimator was marking out observations
that had missing values in the independent variables for
observations outside the subpopulation. This affects the estimated
variance values when the primary sampling units were the individual
observations and could decrease the design degrees of freedom. Both
of these effects are very slight and inversely related to the sample
size. This has been fixed.

6. **svy jackknife** would report too many PSUs when something other than
the default *exp_list* statistic (**_b**) was specified, the PSUs were
data observations, and some of the observations were dropped by the
prefixed command (typically because of missing values), resulting in
slightly inflated degrees of freedom depending on the number of
dropped observations. This has been fixed.

7. **svy: tabulate** with the **se** or **ci** options did not display standard
errors or confidence intervals when there was a stratum with a
single sampling unit, even when the dataset was **svyset** using option
**singleunit(certainty)**, **singleunit(scaled)**, or **singleunit(centered)**.
This has been fixed.

-------- **update 17mar2008** -----------------------------------------------------

__Stata executable, Mac__

1. The permissions for Stata's resource files have been corrected to
allow all users to launch Stata.

2. Some of Stata's dialogs, such as the General Preferences and Find
dialogs, had a close window button mistakenly added to the dialogs
in the 25feb2008 update. This has been fixed.

-------- **update 06mar2008** -----------------------------------------------------

__Ado-files__

1. Following the 25feb2008 update, **xtmelogit** treated a missing value in
the dependent variable as a positive response instead of removing
the observation from the estimation sample. This has been fixed.

-------- **update 03mar2008** -----------------------------------------------------

__Stata executable, Windows__

1. Performance of the Review window has been enhanced so that there is
less overhead between executing multiple commands that are all
pasted into the Command window at once.

2. With the 25feb2008 update, the Do-file Editor overwrote the do-file
in the editor with the selected lines when running those selected
lines using the **Run** button. This has been fixed.

-------- **update 29feb2008** -----------------------------------------------------

__Ado-files__

1. With the 25feb2008 update, the abbreviation **la** for **label** did not
work. This has been fixed.

2. The **predict** dialog reported a "file not found" error when used with
results from the following survey estimation commands: **svy: gnbreg**,
**svy: heckman**, **svy: heckprob**, **svy: intreg**, **svy: logistic**, **svy: logit**,
**svy: mlogit**, **svy: nbreg**, **svy: ologit**, **svy: oprobit**, **svy: poisson**,
**svy: probit**, and **svy: regress**. This has been fixed.

3. The **predict** dialog for **svy: heckman** disabled the radio button for
"Standard error of the prediction" when it should have disabled the
radio button for "Standard error of the forecast". This has been
fixed.

-------- **update 25feb2008** -----------------------------------------------------

__Ado-files__

1. **bootstrap** has the new **jackknifeopts()** option for passing options to
**jackknife** when the **bca** option is specified.

2. The documentation for **glm** and **binreg** did not mention that the
__const__**raints()** option was allowed when fitting models using **ml**
optimization. The help files and dialogs have been updated.

3. **glm**, when used with the **binomial** family and either the **log** or
**identity** link, sometimes issued a spurious warning message regarding
inadmissible predictions. This has been fixed.

4. When using a saved scatterplot matrix that was created using the
**by()** option and that had previously been edited with the Graph
Editor, Observation Property edits might be ignored for some
observations. This has been fixed.

5. The **Y axis** tab has been removed from the dialog box for **graph pie**.

6. **kdensity** displayed a bandwidth of zero on the graph even though the
bandwidth was not exactly zero. **kdensity** now displays more
significant digits for the bandwidth in the resulting graph.

7. **notes** did not attach notes to a variable name with exactly 32
characters. This has been fixed.

8. **predict** with the **scores** option after **mlogit** was displaying an error
when specifying the correct number of variables. This has been
fixed.

9. **predict** with the **scores** option after **ologit** and **oprobit** produced
incorrect equation-level score values when the estimation results
were computed under version control less than 9. This affected the
standard errors produced by **svy: ologit** and **svy: oprobit** when they
were called under version control less than 9. This has been fixed.

10. **ssc whatsnew** and **ssc whatshot** have been renamed **ssc new** and **ssc hot**.
The old subcommand names continue to work but are undocumented.

11. The following changes are made to **sts graph** and **stcurve** for
kernel-based hazard estimation (when option **hazard** is used).

A. The use of the boundary-adjusted kernel functions **epan2**,
**biweight**, or **rectangular** can sometimes lead to negative
estimates of the hazard function. In such cases, the negative
hazard estimates are replaced with 0.

B. When a hazard estimation grid included a time point equal to the
value of a bandwidth, the hazard estimates obtained using an
**epan2**, **biweight**, or **rectangular** kernel were incorrectly shifted
by one time point in the right boundary region. This could
result in a slightly different curve being displayed in the
right boundary region with these kernels. This has been fixed.

C. The left boundary region is changed to be the region [L,L+h)
instead of the old region [0,h). Here L is the minimum analysis
time at which failure occurred, and h is the bandwidth. As
such, the default plotting range is restricted to [L,R] (R is
the observed maximum failure time) for the **epan2**, **biweight**, and
**rectangular** kernels. For other available kernels, the default
plotting range is [L+h,R-h].

12. **svy** results will now report the population and subpopulation sizes
out to a larger number of significant digits, reserving scientific
notation for sizes greater than 99 trillion.

13. **svy: tabulate twoway**, when used with data that contained a stratum
with a single sampling unit, would report zero-valued test
statistics (with p-values of 1) when it should have reported missing
values. This has been fixed.

14. **svy: tabulate twoway** now recognizes value labels assigned to
extended missing values.

15. **svyset** would not check for invalid syntax beyond the stage that
contained an **fpc()** option. This has been fixed.

16. The dialog box for **xtabond** did not allow **vce(robust)** to be specified
with the **twostep** option. This has been fixed.

17. **xtgee** terminated with error 198, Unable to identify sample, if
**pweight**s were used with the **noconstant** option. This has been fixed.

18. **xtgee** did not properly mark the estimation sample (contained in
**e(sample)**) when using the autoregressive, stationary, and
nonstationary correlation structures, and observations were dropped
because they did not have equally spaced time intervals. This has
been fixed.

19. **xtmixed**, **xtmelogit**, and **xtmepoisson** now support time-series
operators.

20. **xtreg, fe** now uses **vce(cluster id)** when **vce(robust)** is specified, in
light of the new results in Stock and Watson,
"Heteroskedasticity-robust standard errors for fixed-effects
panel-data regression," *Econometrica* 76 (2008): 155-174.

__Stata executable, all platforms__

21. The Graph Editor can now record a series of edits, name the
recording, and apply the edits from the recording to other graphs.
You can apply the recorded edits from the **Graph** **Editor** or from the
command line. The edits can be applied from the command line when a
graph is created, when it is used from disk, or whenever it is the
active graph.

See Graph Recorder in **[G-1] graph editor** for creating and playing
recordings in the **Graph** **Editor**. For applying edits from the command
line, see **[G-2] graph play** and the option **play(***recordingname***)** in
**[G-3]** *std_options* and **[G-2] graph use**.

22. Official ado updates may now be distributed as a compressed archive.
This can make updates significantly faster for those with a slow
internet connection.

23. Algorithms for the function for the normal distribution (function
**normal()**) and the log of the normal distribution (function
**lnnormal()**) have been improved to operate in 60% of the time and 75%
of the time respectively, while giving equivalent double-precision
results.

24. **correlate** now returns a matrix **r(C)** containing the correlation or
covariance matrix.

25. **describe** now respects the Results window line width when displaying
variable labels. Long variable labels will wrap to fit the line
width of the Results window.

26. New command **label copy** has been added, which allows a value label to
be copied to a new name.

27. **label values** now takes a *varlist*. To assign the value label, you
specify the label name after the *varlist*. To remove a value label
from a *varlist*, you specify a **.** after the *varlist*. The old syntax
to remove a value label from one variable continues to work.

28. Mata on 64-bit computers now supports matrices larger than 2
gigabytes when the computer has sufficient memory.

29. Mata's **J()** function, **J(***r***,*** c***,*** val***)**, now allows *val* to be specified as
a matrix and creates an *r****rows(***val***)** *x* *c****cols(***val***)** result. The third
argument, *val*, was previously required to be 1 *x* 1. Behavior in the
1 *x* 1 case is unchanged.

30. **merge** would fail to report that the match variables did not uniquely
identify observations in some cases. This has been fixed.

31. **odbc insert** could fail with error r(682) if no previous **odbc**
commands had been executed in the current session. This has been
fixed.

32. **odbc insert** with 64 bit FreeDTS ODBC driver caused an unexpected
error. This has been fixed.

33. Stata will now return an error when it cannot restore data due to
insufficient memory. Stata previously immediately exited. A
**restore** can fail if commands are executed since the last **preserve**
that do not leave sufficient memory to restore the data. You can
free up memory for data by performing a **discard** before a **restore**.

34. The Data Editor could fail to draw data after a graph preference
dialog has been opened. This only affected displaying the data in
the Data Editor; it did not change the data. This has been fixed.

35. **c(changed)** and **r(changed)** are supposed to return a numeric scalar
equal to 0 if the dataset in memory has not changed since it was
last saved and 1 otherwise. However, they incorrectly returned how
many times the dataset in memory has changed since it was last
saved. This has been fixed.

36. Stata would report an "unexpectedly out of memory" error when the **_b**
and **_se** system variables were used within *=exp* expressions that were
parsed by **syntax**. This has been fixed.

37. Some ill-formed xml documents could crash Stata when **xmluse** tried to
read them. This has been fixed.

38. Stata/MP with 16 or more processors and 300 or more variables could
stop with an error message because of mistakenly calculating the
number of processors to use for a certain size of problem. This has
been fixed.

__Stata executable, Windows__

39. Stata updates now have a component for downloading and installing
official utilities. See **update utilities**.

40. Stata now has a mechanism for handling license initialization and
updates under Windows Vista if User Account Control (UAC) is
enabled. This feature requires the above mentioned utilities to be
installed.

41. Programmable dialog boxes now have improved keyboard navigation.

42. Batch jobs were interrupted if Stata's main window was maximized.
This has been fixed.

43. **Ctrl+Shift+S** shortcut for **Save As...** menu item in the Do-file Editor
**File** menu has been added.

44. The Do-file Editor did not handle **\r\r\n** end of line sequences
correctly. This has been fixed.

45. If necessary, the Results window will now maximize before processing
commands from **profile.do**.

46. Pressing **Ctrl+9** crashed Stata. This has been fixed.

47. The background color of the Graph window is now controlled by the
application background color, which is specified in the operating
system's appearance settings. This is typically gray by default.

48. The Graph Editor did not display the extent of certain objects while
dragging. This has been fixed.

49. Using **graph export** to create a TIF image with a width greater than
5300 could cause Stata to crash. This has been fixed.

50. Using **graph export** to create a TIF or PNG image immediately after
exporting to EMF or WMF could produce an image with a black area to
the right of or beneath the graph. This has been fixed.

51. Specifying very small sizes for text in **graph twoway** could cause the
actual size of text rendered on the screen to be larger than
expected. This has been fixed.

__Stata executable, Mac__

52. You may now prevent Stata from notifying you that it has finished
executing a command by bouncing the Stata icon in the Dock. See
**notifyuser** for more information.

Stata will no longer ignore Apple event messages directed at Stata
while it is trying to notify you that Stata has finished executing a
command.

53. Stata now includes Mac OS 10.5 Quick Look support for do-files and
ado-files.

54. Stata can now use Growl (http://growl.info/about.php) to display
notifications that Stata has finished executing a command when it is
in the background. If Growl is installed, Stata displays a short
description of the command and the time that the command has been
completed.

55. The main Stata toolbar now recalls its size from the last time Stata
was used.

56. When running Stata in Mac OS X 10.5 (Leopard), clicking the Review
or Variables window while a Viewer is visible can cause the Viewer
to display the wrong font when the Viewer is later scrolled. This
has been fixed.

57. Specifying very small sizes for text in **graph twoway** could cause the
actual size of text rendered on the screen to be larger than
expected. This has been fixed.

__Stata executable, Unix__

58. The Do-file Editor now encloses filenames with quotes when **do**ing or
**run**ning a file.

59. Pressing the **Control**, **Shift**, or **Alt** keys would dismiss a **--more--**
condition. This has been fixed.

60. Stata could not interpret non-ASCII characters entered into the
Viewer, Data Editor, or Graph Editor. This has been fixed.

61. If the Review window has been sorted by commands, clicking on a
command in the Review window would output the wrong command. This
has been fixed.

62. Using **graph export** to create a TIF image with a width greater than
5300 could cause Stata to crash. This has been fixed.

-------- **update 22jan2008** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 7(4).

2. Robust standard errors would not be computed for **asmprobit**,
**asroprobit**, **asclogit**, and **nlogit** if the model specification would
result in a covariance matrix that had a dimension of 200 or larger.
This has been fixed.

3. **cchart** and **pchart** displayed the wrong number of units out of control
when missing values were present in the first variable. This has
been fixed.

4. **codebook** reported the longest value of a string variable as **str.** if
it was larger than **str100**. This has been fixed.

5. **destring**'s **dpcomma** option did not correctly convert European
formatted numbers when you specified **ignore(".")**. This has been
fixed.

6. **dfuller** after an estimation command overwrote existing e-class
macros. Now previous estimation results are restored unless the
**regress** option is used to display the regression table.

7. New command **estat endogenous**, available after **ivregress 2sls** and
**ivregress gmm**, performs tests to determine whether endogenous
regressors in the previously fitted model can in fact be treated as
exogenous. After GMM estimation, the C (difference-in-Sargan)
statistic is reported. After 2SLS estimation, the Durbin and
Wu-Hausman statistics or robust variants of them are reported.

Also, if a cluster-robust weighting matrix was requested with
**ivregress gmm**, **pweight**s were used, and the **vce()** option was not
specified, a heteroskedasticity-robust VCE was incorrectly reported.
Now, a cluster-robust VCE is reported.

8. The **beta()** option was mistakenly added to the documentation and
dialog boxes for **estat gof** and **estat classification** after **logistic**,
**logit**, and **probit**. The **beta()** option is not allowed, so the help
files and dialog boxes have been fixed to reflect this.

9. **ivregress gmm**, when a HAC VCE was used with two or more lags,
**aweight**s were specified, and the dataset had gaps, would report
incorrect results; this has been fixed. Typically, the differences
between the old and new results are under one percent.

10. Dialog boxes for **lnskew0** and **bcskew0** did not allow the default
confidence level to be calculated. This has been fixed.

11. For large datasets, **lpoly** would sometimes issue **error 3900**, the "out
of memory" error, when computing standard errors. To alleviate this
problem, the code was modified to avoid creating a large temporary
matrix that arises in the intermediate computations.

12. **merge** with the **sort** option previously implied the **unique** option.
**unique** is now implied only if **uniqmaster** or **uniqusing** is not
specified.

13. **mfx** no longer requires you to use the **force** option to obtain the
standard errors of the marginal effects when used after **ivprobit** and
**ivtobit**. Also, the marginal effects with respect to the additional
instruments, which are identically zero, are no longer displayed.

14. **prais** and **newey** did not work with the **by** prefix. This has been
fixed. Also, **newey** would exit with an error message when used with
a dataset that was **tsset** in an earlier version of Stata and no **tsset**
command was issued prior to calling **newey**. This has also been
fixed.

15. **predict**, when used after **nlsur**, would exit with an error message if
the **residual** option was specified and the **equation()** option was not
specified. This has been fixed.

16. Prefix commands that use the **_prefix_command** parsing tool now accept
the **using** clause in the prefixed command. This affects the
following Stata commands: **bootstrap**, **jackknife**, **nestreg**, **permute**,
**rolling**, **simulate**, **statsby**, **stepwise**, and **svy**.

17. qc commands (**cchart**, **pchart**, **rchart**, **xchart**, **shewhart**) now have a
**nograph** option to suppress the display of the graph. These commands
also now return in **r()** the relevant values displayed in the charts.
In addition, **rchart** has the option **generate()** to save the variables
plotted in the chart.

18. Under Windows, **saveold** would return an error when trying to save to
a UNC path (that is, \\computername\path...). This has been fixed.

19. **sktest** now returns a matrix of the test results, **r(Utest)**.

20. **ssc whatshot** did not work when **set varabbrev** had been turned off.
This has been fixed.

21. **svy** regression model commands were marking out observations that had
missing values in the independent variables even for observations
outside the subpop. These observations should be considered part of
the estimation sample for subpopulation variance estimation. This
has been fixed.

22. **xtgls** would not report the log-likelihood or save **e(ll)** when
homoskedasticity and no correlation were assumed, even though the
resulting model is maximum likelihood. This has been fixed.

-------- **update 13nov2007** -----------------------------------------------------

__Ado-files__

1. New command **ssc whatshot** lists the most popular user-written
packages available for download at the SSC. By default, the top ten
packages are listed. Typing **ssc whatshot, n(15)** would list the top
fifteen. (SSC is the premier Stata download site for user-written
software on the web. Additions to Stata available there range from
spectacular to idiosyncratic. If you have not looked at what is
available on the SSC, you should.)

2. For commands **nl** and **nlsur**, when a substitutable expression included
the syntax for a linear combination of variables and a minus sign
preceded that linear combination, the minus sign would be applied to
only the first variable's coefficient. This caused the remaining
variables' coefficients to be displayed with the wrong sign but did
not affect any other results. This has been fixed.

__Stata executable, Mac__

3. Text that has been copied and pasted in the Do-file Editor could
appear incorrectly formatted. A do-file created in Windows and
opened in the Do-file Editor would have a blank line appended after
every line in the do-file. These problems have been fixed.

4. On Mac OS X 10.5, selecting a menu item from the Graph Editor's
contextual menu could lead to unexpected results. This has been
fixed.

-------- **update 29oct2007** -----------------------------------------------------

__Ado-files__

1. **codebook** failed when variable labels started with a right single
quote or had a left single quote within the variable label. This
has been fixed.

2. **fracpoly** and **mfp** have been updated to now work with **intreg**.

3. **stphplot** did not allow version 9 syntax even when called under
version control. This has been fixed.

4. **xtpoisson** with time-series operators produced poorly labeled output.
This has been fixed.

__Stata executable, Mac__

5. The Do-file Editor might crash while executing selected text. This
has been fixed. The horizontal scrollbar might fail to completely
scroll in text from the left edge of the window when using a scroll
wheel that supports horizontal scrolling. This has been fixed.

-------- **update 15oct2007** -----------------------------------------------------

__Ado-files__

1. Dialog boxes that accept **constraints** now provide a tool for managing
them.

2. **destring** now has a **dpcomma** option for handling variables with commas
as decimal points.

3. **egen** now displays a better error message if the **egen** function you
specify does not exist.

4. Marginal effects computed by **estat mfx** after **asmprobit** and
**asroprobit** were incorrect if **estat mfx** computed the means or medians
of nonbinary case variables and the data were unbalanced (not all
cases having the same alternatives). The severity of the error
increased with the degree of alternative imbalance. This has been
fixed.

5. Marginal effects computed by **estat mfx** after **asmprobit** and
**asroprobit** failed if **intmethod(random)** was used during estimation.
This has been fixed.

6. Using the **graph editor** to edit a bar label of a bar graph after
using the editor to change graph orientation, add bar labels,
reverse the scale, stack bars, include missing categories, switch to
dot plot, graph percentages, include 0 in y axis, change bar
spacing, or sort bars produced a graph that could not be used after
being saved to disk as a .gph file. This has been fixed.

7. When adding text to a graph using the Graph Editor's Text Tool, the
properties of the tool could not be changed once the tool had been
used to add text. This has been fixed.

8. **graph matrix** did not allow apostrophes in titles, subtitles, notes,
or captions. This has been fixed.

9. **icd9**'s database has been updated to use codes through the V25 update
of Oct. 1, 2007.

10. The fix to scores after **ivprobit** and **ivtobit** that was made on
30Aug2007 had an unreported implication. Prior to that change, when
**ivprobit** or **ivtobit** were run on survey data, the reported standard
errors were incorrect. This was fixed as of the 30Aug2007 update.

11. **ivregress** could choose the wrong number of lags when a HAC VCE with
automatic bandwidth selection was requested. This has been fixed.

12. The maximum likelihood version of **ivtobit** reported incorrect results
when the censoring point was nonzero or when both left- and
right-censoring points were specified. This has been fixed.

13. Mata's **optimize()** function with a type **v0** evaluator and the **bhhh**
technique would report a conformability error and terminate
execution. This has been fixed.

14. Mata's **optimize()** function with type **v1debug** evaluators reported
that the evaluator did not compute a Hessian matrix when it should
have just ended the derivative-comparison report. This has been
fixed.

15. **ml model** with the **svy** option neglected to mark out observations
where only the survey design variables contained missing values.
This has been fixed.

16. **nlogit** would drop cases that had more than 100 alternatives. This
has been fixed.

17. **predict** with the **stdplna** option after **gnbreg** created a variable
containing the predicted number of events instead of the standard
error of the predicted ln(alpha). This has been fixed.

18. **predict** with the **scores** option after **stcox** reported the error
"option none not allowed" when there were no tied failure times.
This had been fixed.

19. **predict** with the **scores** option is no longer allowed with **stcox**
results that used the **tvc()** option. The **scores** option now requires
that the data be **stsplit** so that the extra observations may be used
to fully generate the requested partial efficient score residuals.

20. **rreg** reported a "no observations" error when the weights in the
biweight iterations were all set to zero. Now **rreg** reports that all
the weights were set to zero in the error message.

21. **stcox** will no longer allow option **tvc()** to be combined with any of
the following variable-generating functions: **basechazard()**,
**basehc()**, **basesurv()**, **effects()**, **esr()**, **mgale()**, **scaledsch()**, and
**schoenfeld()**. These options require that the data be **stsplit** in
order to generate the requested information. The **tvc()** and **texp()**
options are not allowed with the **svy** prefix.

22. **svy: stcox** and **svy: streg** reported a "no observations" error when a
string variable was used in the **id()** option of **stset**. This has been
fixed.

23. **sts graph** options that affect the colors of confidence bands now
apply the exact color specified rather than attenuating the
specified color to a lighter shade.

24. **sts graph** did not respect **ciopts(***recast***(***rarea***))**. This has been
fixed.

25. **svy jackknife** reported incorrect values in **e(b_jk)** when the **mse**
option was not used. This has been fixed.

__Stata executable, all platforms__

26. New **c()** return value **c(tmpdir)** has been added; it returns a string
containing the temporary directory being used by Stata.

27. Options **nrtolerance(***#***)** and **nonrtolerance** have been added to
estimation commands **bprobit**, **blogit**, **cnreg**, **dprobit**, **factor**, **logit**,
**mlogit**, **ologit**, **oprobit**, **probit**, **rologit**, **stcox**, and **tobit**. The
default is **nrtolerance(1e-5)**. Moreover, the iteration log will
display "(backed up)" whenever more than 5 step halves have been
taken to complete a step.

28. Mata's **order()** function allowed up to 500 columns of a matrix to be
used for generating a permutation vector. This limit has been
removed; thus, more than 500 columns of a matrix can be used for
generating a permutation vector.

29. Mata's **sort()** and **_sort()** functions sorted the rows of a matrix
based on up to 500 of its columns. This limit has been removed;
thus, a matrix can be sorted based on more than 500 of its columns.

30. The default amount of memory allocated to ado files was increased
from 500 kilobytes to 1 megabyte.

31. Discrete choice programs **asclogit**, **asmprobit**, **asroprobit**, and **nlogit**
did not drop the temporary variables that encode the case-specific
(cs) variables crossed with alternative indicators that were
collinear, nor did it drop alternative-specific variables that were
collinear with these cs temporary variables. This has been fixed.

32. **expoisson** now scales the exposure or offset variable to help prevent
numerical overflow or underflow from occurring when computing the
relative weights associated with the sufficient statistics.

33. **logistic** displayed untransformed confidence limits in the table of
estimated coefficients when the exponentiated coefficients (odds
ratios) resulted in missing values. **logistic** now reports missing
confidence limits in this case.

34. **mlogit** could crash Stata/MP if the number of variables times the
number of outcomes exceeded matsize. This has been fixed.

35. **odbc** read in BIGINT storage types as LONGs. It now reads in BIGINTs
as strings.

36. **odbc** truncated strings longer than 243 characters to 243 characters.
This has been fixed.

37. Date/time variables were not formatted correctly if you specified a
varlist with **odbc load**. This has been fixed.

38. **regress** with **iweight**s that sum to a value less than one would crash
Stata. This has been fixed.

39. The **stata hidden** command in programmable dialog boxes now respects
version number so that if version is less than 10 the command
behaves as it did in Stata 9, which is a potentially dangerous mix
of the current **stata** **hidden** **immediate** and **stata** **hidden** **queue** that
immediately executes the command while continuing to poll. See
programmable dialog for details on **stata** **hidden**.

40. The maximum table dimension for Fisher's exact test, **tabulate,**
**exact**, is 20 for Stata/IC and 80 for Stata/SE and Stata/MP. These
are the column limits for **tabulate twoway** (see limits).

41. An undetected integer overflow in the algorithm implementing
Fisher's exact test was possible in rare cases involving large
tables (see **tabulate twoway**). In this situation, the algorithm
failed, reporting 0 enumerations at all stages of the enumeration
log, and reporting a p-value of 0. This has been fixed.

42. **translator set pagesize** did not set the correct page height or width
for the page type (legal, a4, etc.). This has been fixed.

__Stata executable, Windows__

43. **Change Working Directory** has been added to the **File** menu making it
consistent with the Unix and Mac versions.

44. **set autotabgraphs on** did not have an effect with floating windows
enabled. This has been fixed.

45. In certain circumstances with floating windows enabled, dockable
windows could become unresponsive. This has been fixed.

46. The Graph Editor did not allow switching between tabbed graphs while
in edit mode. This has been fixed.

47. Using the Graph Editor with the main Stata window minimized would
produce debug output in the Results window. This has been fixed.

48. Saving the contents of the Review window to a do-file would create a
file with mixed end-of-line characters (\r\r\n). In Stata 10, the
Do-file Editor will not handle this sequence. The Review window now
does not write do-files with mixed end-of-line characters.

49. The tooltip for the "Run" button did not change to "Run Selected
Lines" when lines were selected in the Do-file Editor. This has
been fixed.

50. The tab navigation key, the "Enter" key, and the "Esc" key were not
working in the Find/Replace dialog of the Do-file Editor. This has
been fixed.

51. "Replace All" was not reporting the number of changes being made in
the Find/Replace dialog of the Do-file Editor. This has been fixed.

52. The naming convention for a new do-file is "Untitled%d". The number
was not unique among different do-file editors. This has been
fixed.

53. The list of recent files in the Do-file Editor was not saved in the
Windows registry. This has been fixed.

54. The main window was displayed a few pixels outside of the screen
after it was maximized in the previous Stata session. This has been
fixed.

55. (Windows Vista) Under certain circumstances, File dialogs launched
by Stata did not render correctly. This has been fixed.

__Stata executable, Mac__

56. The Do-file Editor's Find/Replace dialog will now stay visible until
the dialog is canceled. A button named "Replace and Find" has been
added that replaces the search text then finds the next occurrence
of the search text.

57. **window manage prefs** {**load** | **save**} has been added. See **window manage**
for more information.

58. The variables dropdown list control from Stata's programmable
dialogs now supports smooth scrolling devices such as Apple's Mighty
Mouse and MacBook trackpads for scrolling. The control will also
display variable names with font smoothing.

59. Ctrl-D has been added as a keyboard shortcut for **File** > **Do**...

60. The Variables window can no longer be sorted by clicking the Format
column header.

61. Saving graphs in the PICT format using **graph export** or the Save
dialog would cause Stata to crash. The **fontface()** option had no
effect on graphs exported in the PICT format using **graph export**.
Both problems have been fixed.

62. Dragging and dropping a graph will now export a graph in the PICT
format if the Copy as PICT general preference is enabled.

63. The Do-file Editor will no longer display the Save dialog to save a
file that was opened by using the **doedit** command.

__Stata executable, Unix__

64. The Do-file Editor's Find/Replace dialog will now stay visible until
the dialog is canceled. A button named "Replace and Find" has been
added that replaces the search text then finds the next occurrence
of the search text.

65. **window manage prefs** {**load** | **save**} has been added. See **window manage**
for more information.

66. Programmable dialog boxes now attempt to scale based on the window
manager's selected font.

67. The Do-file Editor was saving do-files with UTF-8 rather than with a
Latin-1 text encoding. This could cause mistakes in alignment of
output when attempting to run or do the do-file from Stata if the
do-file contained non-ASCII characters. This could also cause
incorrect results in determining string lengths if the string
contained non-ASCII characters. This did not affect data or
numerical results and has been fixed.

68. The Do-file Editor will no longer display the Save dialog to save a
file that was opened by using the **doedit** command.

69. Stata will now bring a dialog back to the front of the main Stata
window after you click on the Variables window to enter a variable
name into the dialog.

70. The Command window was passing commands entered into Stata with
UTF-8 rather than Latin-1 text encoding. This has been fixed.

71. Stata would cause GDK warning messages to appear if a command was
entered into the Command window that contained non-ASCII characters.
This has been fixed.

-------- **update 19sep2007** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 7(3).

-------- **update 30aug2007** -----------------------------------------------------

__Ado-files__

1. **bootstrap** failed to check for **stset** weights when the prefixed
command had the "st" program property. This has been fixed.

2. **codebook** with the **compact** option could alter the sort order of the
data. This has been fixed.

3. **corr2data** now requires the number of rows and columns in the matrix
specified in the **corr()** and **cov()** options be equal to the number of
new variables given.

4. Dialog boxes for **estat** after **arch** and **arima** did not exist. This has
been fixed.

5. **exlogistic** and **expoisson** no longer restrict the **memory(***#***)** option to
a maximum of 512 megabytes.

6. Changes made on the **Organization** tab of the **Legend** **properties** dialog
in the Graph Editor no longer reset the label text for each key to
its original value. Now whether you change a label's text before a
reorganization or after a reorganization the resulting legend is the
same.

7. When using the Graph Editor to edit bar graphs, edits to the text of
bar labels or changes to a bar label's **Object-specific** **properties**
were ignored if the graph were saved and then restored. This has
been fixed.

8. When editing time-series graphs, reference lines could not be added
directly to the plotregion by specifying a date. This has been
fixed.

9. **ivregress** exited with an error if the model did not contain any
exogenous regressors or a constant term and the GMM or LIML
estimator was used. This has been fixed. Also, when the **igmm**
option was used, an error message would be printed before the
iteration log; this too has been fixed.

10. The dialog box for **kdensity** did not output the **kernel()** option
correctly. This has been fixed.

11. Mata's **optimize()** function
**optimize_init_singularHmethod(***S***,"m-marquardt")** would sometimes exit
with the error "Hessian is not negative semidefinite" (or "Hessian
is not positive semidefinite" for minimization) when a Hessian
matrix was not concave (convex). This has been fixed.

12. Mata's **optimize_result_V_robust()** function could result in a
nonsymmetric matrix. This has been fixed.

13. **ml** with the **vce(cluster ***clustvar***)** option reported an incorrect value
in **e(rank)** when there were fewer clusters than model coefficients.
This affected the results in **estat ic** and **estimates stats**, resulting
in larger AIC and BIC values than should have been reported. This
has been fixed.

14. Dialog boxes for **nestreg** and **stepwise** could issue unnecessary error
messages. This has been fixed.

15. The **personal** command would fail on Mac platforms when the **dir** option
was specified. This has been fixed.

16. **predict** with the **scores** option produced incorrect scores following
**ivprobit** and **ivtobit**. This has been fixed.

17. **sample** with an **if** qualifier and the **by()** option (or **by** prefix)
failed to keep the correct number of observations in all **by** groups.
This has been fixed.

18. **stepwise** with **streg** and parentheses binding the first predictor
variable caused the "must specify distribution()" error even when
the **distribution()** option was specified. This has been fixed.

19. **xtlogit**, **xtprobit**, **xtcloglog**, **xtintreg**, **xttobit**, and **xtpoisson, re**
**normal** now use stable sorts for internal computations. This makes
their results perfectly reproducible from run to run so long as the
user does not resort the data. Previously, with some particularly
difficult datasets, it was possible for the results to differ
slightly from run to run.

20. The dialog box for **xtreg** failed to generate a command. This has
been fixed.

-------- **update 30jul2007** -----------------------------------------------------

__Stata executable, Unix (GUI)__

1. Some child dialog boxes did not respond to user input following the
25 Jul 2007 update. This occurred only on Unix platforms and
primarily affected some of the dialog boxes in the Graph Editor.
This has been fixed.

-------- **update 25jul2007** -----------------------------------------------------

__Ado-files__

1. **mfx** with a long **at()** list sometimes produced an erroneous warning
message that at-list variables were not found in the estimation.
This has been fixed.

2. **renpfix** now returns a list of the variables it changed in
**r(varlist)**.

3. **sqreg** did not allow quantiles specified beyond two digits of
precision. This has been fixed.

4. In certain circumstances, **sts graph** did not respect the
**aspectratio()** option. This has been fixed.

5. **xtlogit** with either option **fe** or **pa**, and **xtprobit** with option **pa**,
always exited with an error when used under version control with
version less than 10. This has been fixed.

6. **xtreg, fe cluster()** used incorrect model F test numerator degrees of
freedom when the robust coefficient covariance matrix was not
positive definite. This has been fixed.

__Stata executable, all platforms__

7. **matrix dissimilarity** option **dissim()** now has synonym **s2d()** for
consistency with other commands that have similarity measures as
options.

8. Some graph objects in the Graph Editor could not be deleted by
pressing the Delete key. They could be deleted only by selecting
Object > Delete or right-clicking on the object and selecting
Delete. This has been fixed.

9. Child dialogs did not work correctly if an alternate name was
specified when the parent dialog box was launched. This has been
fixed.

10. Dialog boxes with filenames longer than 28 characters could crash
Stata. Filenames longer than 28 characters are therefore no longer
allowed.

11. The Value Label dialog did not allow value labels to contain spaces.
This has been fixed.

12. Stata previously did not update the Variables window when variable
names with the underscore character as the first character were
dropped. This has been fixed.

__Stata executable, Windows__

13. **graph export** using PNG and TIFF did not work if the **width()** or
**height()** option was specified. This has been fixed.

14. When the user clicked in the Variables window, variables were always
pasted at the end of the text in the Command window. This has been
fixed so that variables will be pasted at the position of the
cursor.

15. Under certain circumstances, the right-click context menu for the
Command window did not have Cut, Copy, and Delete enabled when they
should have been. This has been fixed.

16. The area to the right and beneath a Stata graph did not always
render correctly when the graph was initially drawn. This has been
fixed.

17. When used on systems with a secondary display device, Stata could
attempt to position itself on that device even if the device was not
connected the next time Stata was started. This issue affected
primarily laptops that are sometimes connected to a docking station.
This has been fixed.

__Stata executable, Mac__

18. Stata 10 now uses Mac OS X's Pasteboard for copying images from
Stata to the Clipboard. This allows Stata to copy images in the PDF
format. However, some older applications do not yet support the
Pasteboard. Stata can now revert to the old method of copying
images to the Clipboard in the PICT format by using **set copyaspict**
for such applications.

19. The File > Import menu was mistakenly enabled when Stata was modal
and disabled when Stata was not modal. This has been fixed.

20. Printing a preview of a graph drew the graph in the Graph window
rather than the Preview window. This has been fixed.

21. Selecting lines in the Do-file Editor and selecting Do or Run
executed the whole do-file rather than just the lines selected if
the do-file had been saved. This has been fixed.

22. Stata 10 for Mac now looks in the "Library/Application
Support/Stata" subdirectory of your home directory to search for
your personal ado-directory. The console version of Stata 10 for
Mac was still using the old location, your home directory, to search
for your personal ado-directory. This has been fixed.

__Stata executable, Unix__

23. Stata for Unix(GUI) can now copy graphs as a bitmap to the
Clipboard. A Copy Graph menu item in the Edit menu and contextual
menu, as well as a toolbar button, has been added to the Graph
window to allow copying of graphs.

24. Stata for Unix(GUI) could have problems executing properly on setups
with non-U.S. English locales. Stata now sets the locale to the
traditional Unix system behavior within its own process without
affecting the rest of the system.

25. Sixty-four-bit HP Itanium systems had problems running the GUI
version of Stata 10 because of a missing shared library. This has
been fixed.

26. Some Sun Solaris systems had problems running the GUI version of
Stata 10 for Solaris. Stata 10 has been rebuilt with libraries that
offer better compatibility with various Solaris systems.

27. On Solaris machines with more than 255 fonts, Stata reported that it
could not open the default font configuration file, even though the
file existed and the user had the correct permissions. This has
been fixed.

-------- **update 12jul2007** -----------------------------------------------------

__Ado-files__

1. **arch** now allows you to fit models assuming that the disturbances
follow Student's t distribution or the generalized error
distribution, as well as the Gaussian (normal) distribution.
Specify which distribution to use with the **distribution()** option.
You can specify the shape or degree-of-freedom parameter, or you can
let **arch** estimate it along with the other parameters of the model.

2. **arch** stored an incorrect p-value in **e(p)**, though this did not affect
the displayed results. This has been fixed.

3. **asmprobit** and **asroprobit** did not include the case variable when
setting the estimation-data signature. This has been fixed.

4. **exlogistic** no longer reports p-values for joint tests (as specified
in option **terms()**) using the default method of sufficient
statistics. They were inappropriate because in multiple dimensions
there is no natural ordering of the sufficient statistics. Add
option **test(score)** or option **test(probability)** to obtain p-values
using alternate methods. This can be done either at estimation or
on replay.

-------- **update 03jul2007** -----------------------------------------------------

__Ado-files__

1. Online help and the search index have been brought up to date for
Stata Journal 7(2).

2. Edits made in the graph editor may now be undone or redone any time
in the current Stata session, even if the editor is stopped and then
restarted.

3. The indicator that a graph has been edited in the graph editor now
turns off when all edits have been undone.

4. **estat firststage** after **ivregress** exited with an error when the model
contained time-series operators. This has been fixed.

5. **nlogitgen** did not allow the alternatives variable to be of type
string. Now it does. The new alternatives variable generated by
**nlogitgen** is now generated as an **int** instead of a **float**.

6. The variable **trunc** automatically created by **uselabel** could contain
missing (**.**) when it should contain **0** if **uselabel** encountered more
than 100 label mappings. This has been fixed.

-------- **update 26jun2007** -----------------------------------------------------

__Stata executable, all platforms__

1. **exlogistic**, in the unusual case where the problem was large and the
**memory()** option was set to its maximum of **512m**, could crash Stata.
This has been fixed.

2. Multiple new-line characters in a string could cause Mata's **printf()**
function to enter an endless loop. This has been fixed.

3. On very-high-resolution displays, the Data Editor could not display
more than 20 columns at a time. This has been fixed.

__Stata executable, Windows__

4. Exporting graphs as PNG or TIF files could produce an image with a
black area to the right of or beneath the graph. This has been
fixed.

5. The file dialog for saving commands from the Review Window could
fail to append the file extension. This has been fixed.

__Stata executable, Mac__

6. The Variables window would show temporary variables that no longer
exist after some commands. This has been fixed.

-------- **previous updates** -----------------------------------------------------

See whatsnew9to10.

-------------------------------------------------------------------------------