Stata 15 help for whatsnew12

Additions made to Stata during version 12

This file records the additions and fixes made to Stata during the 12.0 and 12.1 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 | | this file 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 | | whatsnew10 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 whatsnew12to13.

-------- update 23jan2014 -----------------------------------------------------

1. The 25nov2013 update introduced a bug where you could not add an object of default size to the SEM Builder by simply clicking where you wanted the object placed in the SEM Builder. You had to click and drag to add and set the size of the object. This has been fixed.

-------- update 18dec2013 -----------------------------------------------------

1. ivregress, when specified with interactions involving factor variables with large numbers of categories, produced an error message. This has been fixed.

2. nlcom and lincom, when estimating an expression involving a comma (like _b[/cov(e.x,e.y)]), produced an error message. This has been fixed.

-------- update 25nov2013 -----------------------------------------------------

1. asmprobit failed to display the coefficient table when there were no case variables and no constant terms in the model. This has been fixed.

2. Business calendars use dateformat to specify how dates are typed in business calendar files. The options for dateformat are ymd, ydm, myd, mdy, dym, and dmy. Not all of these options worked, for example, in function bofd() or command bcal describe. This has been fixed.

3. cluster with subcommands averagelinkage, completelinkage, waveragelinkage, medianlinkage, centroidlinkage, or wardslinkage and with a large number of observations (such as 400,000) could crash Stata instead of issuing an error message. This has been fixed.

4. Function dunnettprob() would return a missing value if the df argument was not an integer value. dunnettprob() now behaves like invdunnettprob() and allows the df argument to be any real value in its documented domain.

5. expoisson, when using a depvar with a large number of counts (total exceeding 125) and an exposure variable, could give wrong results. This has been fixed.

6. export excel did not export a variable formatted as %d as a date. This has been fixed.

7. fcast compute has the following fixes:

a. fcast compute after vec failed to clear a global macro that it uses during execution and would therefore accumulate a long string in this macro when called repeatedly. When the resulting string exceeded the maximum allowed length, the command would produce error "macro substitution results in line that is too long". This has been fixed.

b. fcast compute after vec dropped all variables created by previous calls to fcast compute if the current call resulted in an error. This has been fixed.

8. gmm, when option xtinstruments() was specified and the user requested a cluster-robust weight matrix or VCE, ignored the cluster variable and instead clustered based on the panel variable. This has been fixed.

9. The icd9 databases have been updated to use the latest version (V31) of the ICD-9 codes.

10. For .xls files, import excel might omit rows or columns at the end of the sheet because of the recorded number of rows or columns in the .xls file not matching the number of actual rows or columns. This has been fixed.

11. ivregress has the following fixes:

a. ivregress 2sls is now more tolerant of highly collinear regressors.

b. ivregress with option first would report an error when factor variables were included in the exogenous variable list that were highly correlated with the instruments. This has been fixed.

c. ivregress would report a syntax error when a very long if expression was used. This has been fixed.

d. Option noconstant was not correctly passed through the collinearity check of the instruments. This has been fixed.

12. Mata function st_addvar() with more than one variable type specified and with the number of variable types fewer than the number of variable names could crash Stata. This has been fixed.

13. Mata function asarray_keys() generated a 3201 error when there were collisions of the hashes. This has been fixed.

14. Mata function st_view() is now much faster at creating views when the variable specification does not contain time-series or factor-variable operators.

15. matrix accum would incorrectly compute the cross-product sum between a continuous variable and a factor variable when interactions involving continuous variables were also specified. This has been fixed.

16. mi has the following fixes:

a. mi estimate produced an error if the length of the specified command line exceeded 244 characters. This has been fixed.

b. mi estimate, when used with mean, proportion, ratio, or total, could produce missing multiple-imputation standard errors when some of the imputation-specific standard errors were legitimate zeros. For example, the mean command reports a standard error of zero for a constant variable. mi estimate now reports the estimates instead of missing values.

c. mi estimate used with mean, proportion, ratio, or total now produces missing multiple-imputation standard errors for any parameter that was estimated based on a sample of only one observation in any of the imputations, that is, for any parameter for which the corresponding element of the e(_N) matrix is one in any of the imputations.

d. mi impute chained would mistakenly produce error "missing imputed values are produced" when option noimputed was specified and option include() contained a categorical or binary imputation variable with a factor specification other than i.. This has been fixed.

e. During conditional imputation, mi impute chained and mi impute monotone failed to check the order in which the conditional and conditioning variables were imputed when the conditional variable contained the same number of missing observations as one of the conditioning variables. The commands now produce an error if a conditional variable is listed before its conditioning variables in the specification of an imputation model.

f. mi impute mvn, when the check for collinear independent variables failed, produced an error with a Mata traceback log. It now produces an appropriate error message.

17. ml with option subpop() reported an incorrect sample size when some of the variables contained missing values outside the subpopulation sample. It was also possible for an entire sampling unit to be dropped because of these missing values, yielding subpopulation variance estimates based on an incorrect sample size. This has been fixed.

18. nestreg with prefix svy and svy options exited with a syntax error when the prefixed estimation command contained parentheses-bound variables. This has been fixed.

19. nlogit could drop too many alternative-specific variables if one alternative-specific variable was collinear with a variable specified in the last (leaf) alternative varlist after it was interacted with indicators of the alternatives. This has been fixed.

20. The dialog box for predict after arfima did not produce a command when you pressed the Submit button. This has been fixed.

21. In rare situations, qreg might cause Stata to crash. This has been fixed.

22. The parameter domains for rnbinomial(), the negative binomial random-number generator, have changed to allow a greater range of parameter combinations.

23. Performing a search in Stata's help system while Stata was busy executing a command could cause the command to fail with an error or even cause Stata to crash. Stata no longer allows a search in Stata's help system while Stata is busy executing a command.

24. set cformat, set pformat, and set sformat have the following fixes:

a. They now report an error when the specified format width is too large. The maximum format width for cformat is 9. The maximum format width for pformat is 5. The maximum format width for sformat is 8.

b. They reported error "invalid numeric format" on an empty format specification when option permanently was specified with a space before the comma. This has been fixed.

25. set obs, when a variable was sorted and when the last value of that variable contained an extended missing value, did not mark that variable as no longer being sorted. This has been fixed.

26. When the jackknife method was used, stptime and strate could report jackknife confidence intervals that did not include the estimated rate for survival datasets that contained no failures, only one failure, or only one cluster with failures. They now report missing confidence intervals and a note with an explanation in these situations.

27. ivregress with prefix svy has the following fixes:

a. Interactions in the exogenous variable list caused an error. This has been fixed.

b. When svy: ivregress was used with option first, factor variables in the exogenous variable list caused an error. This has been fixed.

28. svy tabulate, when used with a string variable, reported a temporary variable name in the output instead of the specified variable. This has been fixed.

29. symmetry with an empty if condition returned a tempvar not found error. This has been fixed.

30. Function tukeyprob() would return a missing value if argument df was not an integer value. tukeyprob() now behaves like invtukeyprob() and allows df to be any real value in its documented domain.

31. var with option small and constraints would display missing values for the F statistic in the header. This has been fixed.

32. xtmixed with collinear variables in a random-effects equation would exit with an error and a Mata trace log. This has been fixed.

33. Estimation commands that allow factor-variable notation would sometimes spuriously omit a continuous variable because of collinearity when interactions involving continuous variables were also specified. This has been fixed.

34. When executing an ado-file that was defined with more than 3,500 lines, Stata could crash. Now Stata correctly issues the error message "system limit exceeded".

35. Stata would appear to lock up while executing a do-file when there was an I/O error while reading the do-file (that is, the do-file was erased while Stata was executing it). Stata now stops and displays an error message when there is an I/O error while executing a do-file.

36. The Data Editor has the following fixes:

a. Long value labels could overflow the width of a column, causing them to display in adjacent columns. This has been fixed.

b. Pasting data into the Data Editor did not update the internal list of sorted variables. Now if you paste data into one of the variables on which the data were sorted, the list is updated.

37. The SEM Builder no longer allows you to add an object that is of 0 width or height.

38. Syntax highlighting for the global macro in the Do-file Editor passed beyond the white-space character. This has been fixed.

39. (Stata/MP) matrix score, when used with an if expression that could not be run in parallel (for example, if runiform()>0.5), might produce incorrect results. This has been fixed.

40. (Windows) file seek always failed with files larger than 2 gigabytes. This has been fixed.

41. (Windows) graph export would fail with error r(603) when used with a UNC path. This has been fixed.

42. (Windows) set httpproxyport would accept ports between 1 and 32,000, but a setting with ports above 9,999 would not be reloaded in subsequent Stata sessions. This has been fixed.

43. (Windows) The Do-file Editor Preferences dialog set the wrong default font size when button Restore Factory Defaults was pressed from the Editor Font tab. This has been fixed.

44. (Windows) update failed to install from a local directory when Stata was run from a network share. This has been fixed.

45. (Mac) The Do-file Editor now uses a different method for saving temporary files to disk before running them. The previous method prevented data loss in case of a crash if a file was being overwritten. However, this method was unnecessary for temporary files and could make a temporary file unavailable when Stata was ready to run it.

46. (Mac) import excel and export excel did not convert strings from Unicode when importing and to Unicode when exporting. This has been fixed.

47. (Mac) odbc load failed to load variables of type SQL_WCHAR, SQL_WVARCHAR, and SQL_WLONGVARCHAR on Macs. This has been fixed.

48. (Mac) Saved preference max_memory would not be read correctly if the setting was large. This has been fixed.

49. (Mac) Mac OS X 10.8.2 introduced a bug that could cause Adobe Reader to not receive messages from Stata to open the Stata PDF Documentation. This could happen if the user quit Adobe Reader and later clicked on the Stata PDF Documentation link from within Stata. Stata now has a work-around for this bug.

50. (Mac) Point marker symbols in graphs that are exported to PDF did not appear because the marker symbols were too small. This has been fixed. The marker symbol is now rendered as a circle rather than a square.

51. (Mac) Graphs could not be exported to EPS or PS while being edited in the Graph Editor. This has been fixed.

52. (Mac) Stata could crash if multiple graphs were drawn while the Do-file Editor was in full-screen mode. This has been fixed.

53. (Mac) The Retina version of the SEM Builder's tools icons was not being used by Mac OS X. This has been fixed.

54. (Mac) Stata displays a warning when the Data Editor is changed from browse mode to edit mode. If data changed and you typed edit in the Command window while in browse mode, Stata would display a warning dialog and immediately dismiss it. This has been fixed, and the warning dialog is no longer immediately dismissed.

55. (Mac) Selecting Edit > Undo after editing text in the Data Editor and applying the change would cause Stata to crash. This has been fixed.

56. (Mac) Using translate to convert a SMCL file to a PDF file would not produce a PDF file if the PDF filename contained spaces. This has been fixed.

57. (Mac) The notification sound Stata plays when Stata is finished executing a command in the background would not play in OS X 10.7 and 10.8. This has been fixed.

58. (Mac) OS X 10.9 (Mavericks) introduced a change that caused Stata to always start in the root directory "/" instead of the working directory from Stata's previous session. This would only affect Stata if it was launched by double-clicking on it in the Finder, not by double-clicking on a do-file or dataset from the Finder. This has been fixed.

59. (Unix) Stata for Unix GUI crashed when you selected the menu item Edit > Preferences > Manage Preferences > Factor Settings with a Viewer open. This has been fixed.

-------- update 09jul2013 -----------------------------------------------------

1. histogram incorrectly reported the syntax error "parentheses do not balance" when a suboption of by() was specified with quoted spaces. This has been fixed.

2. intreg without option het() left behind an invalid e(V_modelbased) when vce(robust) or vce(cluster ...) was specified. This only affected margins when option vce(unconditional) was specified. This has been fixed.

3. margins has the following fixes:

a. margins with option vce(unconditional) used after xtmixed would produce a cryptic error about not finding an equation. This has been fixed.

b. margins would report the warning "cannot perform check for estimable functions" with reg3 results that did not contain constraints for factor variables. This has been fixed.

c. margins was not properly handling quoted strings in option predict(), causing an error while reporting results. This has been fixed.

4. marginsplot produced an error message or a plot with missing points when margins was specified with option at() that set a value out of the range for a given regressor. This was most common when the at() variable was of type byte and the at() value was greater than 100. This has been fixed.

5. Mata function moptimize_result_display() with the Nelder-Mead technique and constraints would report "estimates post: matrix has missing values" when it should have reported the estimates table with missing standard errors. This has been fixed.

6. ml plot and ml graph, when specified to produce as many or more plotted points than observations in the dataset, would exit with a Mata trace log when these commands should have produced a graph. This has been fixed.

7. pwmean would interact only the first two variables specified in option over() when it should have been interacting all the over() variables. This has been fixed.

-------- update 20mar2013 -----------------------------------------------------

1. Online help and the search index have been brought up to date for Stata Journal 13(1).

2. predict, residuals after areg for models using time-series operators would fail with the error "not sorted". This has been fixed.

3. In some cases, proportion returned a cryptic error message when it should have returned "no observations". This has been fixed.

4. tsset with a time variable in business calendar (%tb) format reported an incorrect time period (delta). This has been fixed.

-------- update 25feb2013 -----------------------------------------------------

1. sem with method(mlmv) will now attempt to fit the specified model even if the saturated model fails to converge.

2. xtile produced missing quantiles when an if condition was used and the resulting number of observations was less than the requested number of quantiles. It now produces an error in this case.

-------- update 30jan2013 -----------------------------------------------------

1. expandcl, cluster() generate(), when there were many clusters before expanding the data, failed to generate unique identifiers for the duplicated clusters. This has been fixed.

2. graph with option yscale(range()) did not properly expand the range when the natural range of the data was negative and when the maximum value specified in range() was also negative. This has been fixed.

3. gmm would sometimes exit with an error message stating that "panels are not nested within clusters" when panel-style instruments were specified with option vce(cluster id) or wmatrix(cluster id), where id is the panel variable previously set via xtset. This has been fixed.

4. ivtobit would produce incorrect results if an endogenous regressor was adorned by time-series operators. This has been fixed.

5. In the SEM Builder, moving text using the position control in the text properties dialog failed. This has been fixed.

6. stcox and streg with option shared() no longer allow the presence of delayed entries (left truncation) or gaps (interval truncation). In general, the use of this option for truncated survival data will lead to inconsistent results unless the frailty distribution is independent of the covariates and the truncation point, which is a rather restrictive assumption in practice. If desired, the estimation can be forced in the described case by specifying the undocumented option forceshared, although we strongly discourage users from doing so.

-------- update 18dec2012 -----------------------------------------------------

1. infile (free format), in the rare case where the end-of-file designation was encountered in the middle of an observation, would mistakenly drop that entire observation and fill in the remaining variables in the previous observation with missing value . or "". This has been fixed.

2. ivprobit's maximum likelihood estimator reported the correct p-value for the exogeneity Wald test but stored the incorrect value in e(p_exog). This has been fixed.

3. Two updates were made to margins:

a. margins no longer allows interaction terms to contain continuous variables in the margins list.

b. margins gave the error "not an array" if there was a Stata graph in memory with one of the following names: asbal, base, mstat, phat, or p. This has been fixed.

4. matrix accum with option deviations and factor-variable notation indicating a subset of the levels of a factor variable interacted with at least one continuous variable would produce incorrect values for off-diagonal elements of the cross product with other nonfactor terms. This has been fixed.

5. ml using evaluator type gf0 with constraints and option vce(robust) or vce(cluster ...) would report a Mata trace log after converging. This has been fixed.

6. ml check would report a Mata trace log when multiple techniques were specified using option technique() of ml model. ml check now reports an error if anything other than technique(nr) is in effect.

7. net would not show all the Stata packages available from a particular location if there were more than 150 entries in the table of contents file for that location. This has been fixed.

8. notes search failed to find the text that was searched for if no dataset-wide (_dta) notes existed. This has been fixed.

9. Several updates were made to the SEM Builder:

a. You can now easily change the confidence level used to compute confidence intervals through the new menu item View > Confidence Level. Changes made there are reflected in the Confidence level field on the Reporting tab of the SEM estimation dialog. Likewise, changes on the dialog are reflected in View > Confidence Level.

b. The typeface of results displayed on paths and covariances could not be changed and the italic and bold settings were ignored. This has been fixed.

c. The confidence intervals for variances displayed in the Properties pane were computed in the variance metric, meaning that the lower bound could become negative. While asymptotically equivalent, these CIs did not match those shown on the diagram and in the Results window. These CIs are now computed in the log variance metric and transformed back into the variance metric. They now match the CIs on the diagram and in the Results window.

d. Settings made in the SEM Builder are now remembered for all future path diagrams, even those drawn in ensuing Stata sessions. Setting the label on a variable to "None" in a settings dialog correctly removed the label from the current diagram, but the setting was not remembered when new diagrams were created in new SEM Builder windows. This has been fixed.

10. Several updates were made to sem:

a. sem with suboption pattern() or fixed() of option covstructure() was using the upper triangle elements of the matrix argument instead of the lower triangle elements. sem now uses the lower triangle elements, as was originally intended and documented.

b. sem with option standardized now assumes option showginvariant for models fit with option groups().

c. sem reported a cryptic error message when it encountered a model where an observed endogenous variable had a path to a latent variable but the model did not contain any observed exogenous variables. This has been fixed.

11. xtdpd and xtabond, when options twostep and vce(robust) were both specified and when the number of panels was small relative to the number of instruments, could result in unstable standard-error computations and could lead to standard errors with potentially poor coverage. This was rare but was most likely to occur when the number of instruments exceeded the number of panels. The stability of computations has been dramatically improved, and the new computations have nominal coverage.

12. (Mac) When pasting text into the Data Editor, any text that could not be converted to ASCII would result in no text being pasted. Stata will now display a prompt to convert the text using lossy conversion, which removes or alters some characters. For example, accented characters that cannot be converted will lose the accent.

13. (Unix) If many preferences were cleared, it was possible though highly unlikely that some of the cleared settings would still be used in future instances of Stata. This has been fixed.

14. Online help and the search index have been brought up to date for Stata Journal 12(4).

-------- update 22oct2012 -----------------------------------------------------

1. The path diagrams shown in the examples in [SEM] Stata Structural Equation Modeling Reference Manual can now be opened in the SEM Builder. In the Builder, click on the Help menu and select Example Diagrams.

2. gmm exited with an error message stating an equation had no parameters if a substitutable expression contained only linear combinations declared in previous expressions and no individual parameters. This has been fixed.

3. The icd9 databases have been updated to use the latest version (V30) of the ICD-9 codes, which became effective on 1 October 2012.

4. (Windows) If multiple graphs were tabbed in the same Graph window, Stata could enter an endless loop when graph preferences were changed. This has been fixed.

5. (Windows) Toolbar items and menus that list open graphs did not show all graphs if they were tabbed. This has been fixed.

6. (Mac) Importing an Excel file from the Import Excel dialog caused Stata to crash if the Excel file contained no worksheets. This has been fixed.

7. (Mac) You will no longer be prompted to check for updates after loading a saved preferences set and restarting Stata.

8. (Mac) In the previous update, Stata updated its file and application icons to the format Apple recommends for support by Macs with a Retina Display. However, Macs running Mac OS X 10.5 do not support this new icon format. Stata now uses the older icon format. Stata still uses other graphical assets that are compatible with both Macs with a Retina Display and Mac OS X 10.5.

9. (Mac) If the interface layout was set to Combined and the Variables/Review views were hidden, selecting them from the Window menu still kept them hidden. This has been fixed.

-------- update 01oct2012 -----------------------------------------------------

1. Online help and the search index have been brought up to date for Stata Journal 12(3).

2. graph twoway style options lcolor() and lpattern() could trigger a bug in _parse factor, causing Stata to crash. This has been fixed.

3. import excel gave a "file too big" error for .xls files over 40 MB. The restriction was intended for .xlsx files only. This has been fixed.

4. lpoly would on rare occasions exit with an error message when fweights or aweights were specified. This has been fixed.

5. margins commands with an at() option that contained repeated values for a given variable repeated the margins results. This was not the intended behavior, so now margins keeps only unique values for a given variable.

6. marginsplot failed if the preceding margins command had an at() option with repeated values for a given variable. This has been fixed.

7. marginsplot failed if any of the variables involved in the plot had a value label assigned and that value label contained no labels. This has been fixed.

8. Mata moptimize() subroutines moptimize_util_vecsum() and moptimize_util_matsum() now skip over observations where the score variable is equal to zero. This change was made to facilitate likelihood-evaluator functions that have a method for dealing with missing values in indepvars.

9. matlist when used with factor-variable notation in the column titles would unintentionally show some SMCL tags in the output. This has been fixed.

10. matrix accum with the deviations option and factor-variable notation indicating a subset of the levels of a factor variable interacted with at least one continuous variable would produce incorrect values for off-diagonal elements of the cross product with other terms containing factor variables. This has been fixed.

11. mi impute intreg and mi impute truncreg could sometimes produce imputed values outside the implied range when the set of complete predictors exhibited some collinearity. This has been fixed.

12. odbc insert when inserting data into a dBASE file sometimes produced an error message and failed to produce the file. This has been fixed.

13. predict after arch and arima did not calculate dynamic forecasts correctly when the option dynamic(.) was specified. This has been fixed.

14. reg3 and sureg with a complex system containing many identifying constraints and a relatively small number of observations would sometimes report that the estimated covariance matrix was not symmetric, yielding missing values for unconstrained model coefficients. This has been fixed.

15. _return hold results named by using tempname were not being dropped automatically at the program's conclusion. This has been fixed.

16. _return restore with the hold option did not copy matrix stripes correctly, eventually causing Stata to crash. This has been fixed.

17. In the SEM Builder, holding the Shift key while adding a path or covariance will now constrain them horizontally or vertically.

18. The SEM Builder now displays standardized estimates when "Display standardized coefficients and values" is checked on the Reporting tab of the Builder's SEM estimation options dialog box.

19. spearman no longer creates unnecessary temporary variables.

20. tssmooth ma exited with an error message if the weights() option was specified and the dataset had more than 3,201 observations. This has been fixed.

21. Stata's variable list parse logic was not distributing time-series operators on interaction terms correctly when the time-series operators were specified on a parentheses-bound list containing interaction terms. This has been fixed.

22. (Mac) In the SEM Builder, on the Group tab of the SEM estimation options dialog box, selecting a parameter from the "Parameters that are equal across groups" combobox would insert the textual description of the parameter instead of its option name into the combobox. This would result in an error when submitting the command. This has been fixed.

23. (Mac) On a Mac with a Retina Display, the selected tab of a window that did not have the keyboard focus would render the wrong color. This has been fixed.

24. (Mac) When using the Import Excel dialog, selecting a filename with Unicode characters could cause Stata to crash. This has been fixed.

-------- update 08aug2012 -----------------------------------------------------

1. glm would exit with an error message if one or more levels of a factor-variable term was omitted because of collinearity. This has been fixed.

2. The dialog box for import excel failed to output the correct sheet name if the first sheet was empty. This has been fixed.

3. sem with summary statistics data (ssd) failed to properly compute the log likelihood for the saturated model when one of the observed variable names was a substring in another observed variable name (for example, when some of the variable names were composed of one letter). As a result, sem failed to properly fit the specified model and to report the likelihood-ratio test comparing the fitted model with the saturated model. In addition, estat gof reported a Mata trace log. This has been fixed.

4. (Mac) A change in Mac OS X 10.8 (Mountain Lion) caused Stata to crash when automatically relaunched after a Stata update. Although all ado-files and the Stata executable would be successfully updated, the user had to manually restart Stata after an update. This has been fixed.

-------- update 18jul2012 -----------------------------------------------------

1. alpha now reports a more informative error message when it cannot determine the sense of each item empirically.

2. blogit, bprobit, glogit, and gprobit, with options vce(bootstrap) and vce(jackknife), treated the population total variable, pop_var, as an independent variable when checking for collinearity. This has been fixed.

3. graph twoway function passed the wrong version to the function f(). This rarely mattered but could affect invbinomial(). This has been fixed.

4. If position() was specified in the title_options of legend() in graph twoway, the same position was wrongly applied to the main title of the plot as well. This has been fixed.

5. import excel can now open Excel files directly from a URL.

6. The import excel dialog box disabled the checkbox "Import first row as variable names" if the first row of data contained numeric values. This has been fixed.

7. keep _all would mistakenly drop all variables. This has been fixed.

8. (Windows) When you copy an SEM diagram to the Clipboard, the entire workspace background is now included.

9. (Mac) Line wrapping can now be disabled when printing from the Do-file Editor by using the "Wrap lines" checkbox from the General Preferences dialog and the Do-file Editor > Advanced tab.

10. (Mac) A bad symbolic link (for example, a symbolic link that points to a file that does not exist) can cause Stata's Open dialog to hang requiring a Force Quit of Stata. This has been fixed.

11. (Mac) Stata for Mac is now code signed as required by Mac OS X 10.8 (Mountain Lion).

12. (Ubuntu Linux) Recent releases of Ubuntu Linux (for example, 12.04) use the Linux glibc library version 2.15, which contains a bug in the math function logb(x). This bug affects the commands exlogistic and expoisson. Stata now uses its own implementation of logb() internally and so is no longer affected by the bug in glibc.

13. (Unix) The Data Editor deleted multiple characters after the first character was deleted if you tabbed into a cell. This has been fixed.

-------- update 03jul2012 -----------------------------------------------------

1. Stata has been updated to account for the leap second that was added by the International Earth Rotation and Reference Systems Service on June 30, 2012. Stata's datetime/C UTC date format now recognizes 30jun2012 23:59:60 as a valid time.

-------- update 27jun2012 -----------------------------------------------------

1. Online help and the search index have been brought up to date for Stata Journal 12(2).

2. contrast with weighted contrast operators computed improperly weighted effects. This happened because svy failed to post the proper weighted frequencies of factor variables. This has been fixed.

3. margins failed to compute properly weighted effects when weighted contrast operators were specified and the estimation results used weights. This has been fixed.

4. sem commands built by using the SEM Builder with option method(mlmv) and data with missing values in the estimation sample caused an unintentional error in the following postestimation commands:

estat ginvariant estat mindices estat residuals estat scoretests estat teffects

This has been fixed.

5. If the SEM Builder automatically added many covariance paths during estimation, it sometimes failed to put the estimation results in the path diagram. This has been fixed.

-------- update 20jun2012 -----------------------------------------------------

1. datasignature set, confirm, and report reported a file-not-found error (or saved a signature file without extension .dtasig) when the specified path and filename began with a "." or "..". This has been fixed.

2. egen, when the argument to fcn() was an expression of a variable subtracted or multiplied by another variable, mistakenly unabbreviated the argument as if it were a variable list. This has been fixed.

3. estat teffects after sem would incorrectly label the values for direct, indirect, and total effects and their standard errors for some fitted models that contained one or more exogenous variables and at least two endogenous variables. This has been fixed.

4. marginsplot failed with an error message when Stata's temporary directory contained spaces. This has been fixed.

5. Mata function st_fopen() reported a file-not-found error when the specified path and filename began with a "." or "..". This has been fixed.

6. predict after arima ignored option stdp and produced linear predictions, the default. This has been fixed.

7. sem reported the syntax error "option ... not found" when a display option such as standardized or level() was used with a single path specification that was not bound in parentheses. This has been fixed.

8. tsappend, last() tsfmt() was slow when there were many panels. This has been fixed.

-------- update 23may2012 -----------------------------------------------------

1. New command icc computes intraclass correlation coefficients for one-way random-effects models, two-way random-effects models, and two-way mixed-effects models for both individual and average measurements. Intraclass correlations measuring consistency of agreement or absolute agreement of the measurements may be estimated.

2. New postestimation command estat icc computes intraclass correlations at each nested level for nested random-effects models fit by xtmixed and xtmelogit.

3. estat gof has been disallowed after svy, subpop(): because this test does not specifically address subpopulation estimation.

4. import excel mistakenly imported some Excel formats that included brackets as datetime formats. This has been fixed.

5. Adding a Mata function to a Mata library failed with an error if the function had an optional argument that was a Mata class matrix, a Mata class rowvector, or a Mata class colvector. This has been fixed.

6. odbc insert has new option block used for block inserts. odbc insert now allows you to specify qualifiers if and in.

7. (Mac) Stata for Mac was not using the ISO Latin1 string encoding to output text when set charset was set to latin1. This has been fixed. This setting affects all of Stata's text output, including graphs, the Data Editor, and the Do-file Editor, not just SMCL output in the Results and Viewer windows.

8. (Mac) When using the Find Panel to search and replace all text in the Do-file Editor, Stata could crash if the file being edited contained multiline comments. This has been fixed.

9. (Mac) The Mac OS X 10.7.4 update could cause Stata's help system to crash. This has been fixed.

10. (Mac) Entering accented characters in the Data Editor could cause Stata to crash. This has been fixed.

11. (Windows) In the Variables window, the scrollbar might not show up if the scrollbar was moved from the top position and a new dataset was loaded. This has been fixed.

12. (Unix) Copying data from Excel and pasting the data into the Stata for Unix Data Editor would sometimes cause Stata to crash. This only affected users who accessed Stata on a remote machine with an X server or who used virtualization software to run both Unix and Windows operating systems simultaneously. This has been fixed.

-------- update 11may2012 -----------------------------------------------------

1. egen, when the argument to fcn() was * (all variables), included a temporary variable intended for egen's own use. This has been fixed.

2. estimates store, when the specified name was an abbreviation of a previously specified name, dropped variable e(sample) of the previously stored estimation. This has been fixed.

3. import excel did not read Microsoft Excel files if the file extension was capitalized. This has been fixed.

4. estat gof after logistic, logit, and probit issued a "not enough degrees of freedom to perform the test" error when it was possible to perform the test but there was only one degree of freedom. This has been fixed.

5. fcast compute, when used after command vec with option sindicators(), produced incorrect forecasts. This has been fixed.

6. margins reported a conformability error when used with mlogit or mprobit estimation results that used a replication method for variance estimation, such as with options vce(bootstrap) and vce(jackknife) or with prefix commands bootstrap, jackknife, svy bootstrap, svy brr, svy jackknife, and svy sdr. This has been fixed.

7. In rare cases, the SEM Builder created .stsem files that it could not open. A sequence of four actions prior to saving was required to elicit the behavior. 1) The SEM model had to be fit within the Builder and the diagram had to be incomplete in that some paths or covariances required to identify the model were missing. 2) When presented with a dialog box asking whether to add the missing paths or covariances, the user had to select "Yes". 3) The user then had to modify the placement or appearance of one of the added paths or covariances. 4) The model had to be refit. This has been fixed.

If you have a .stsem file that exhibits this problem, StataCorp can recover the file. Send the file to Technical Services at tech-support@stata.com. If you can also send the associated dataset, the recovery will be easier and quicker, but the dataset is not required.

8. In the SEM Builder, some cosmetic edits to the path diagram (for example, color and size) could not be undone. This has been fixed.

9. sts graph produced an error when the path of the temporary directory contained a space. This has been fixed.

10. tetrachoric with option pw and two variables with no nonmissing observations in common produced a spurious correlation value instead of a missing value for those two variables. This has been fixed.

-------- update 24apr2012 -----------------------------------------------------

1. export excel exported blanks instead of the numeric value for data that used a value label, but the data were unlabeled. This has been fixed.

2. margins now works with constant-only models.

3. pwcompare and pwmean did not account for unequal within-term variances of the margins when determining group codes; thus some significantly different margins were incorrectly labeled with the same group code. This has been fixed.

4. (Stata/MP) regress with aweights that did not sum to the sample size, option noconstant, and 300 or more independent variables computed an incorrect root mean squared error and variance-covariance matrix of the coefficient estimates. This has been fixed.

5. (Mac) The Window > Do-file Editor > New Do-file menu item has been renamed to Top Do-file and assigned a keyboard shortcut. This menu item will select the topmost Do-file Editor window and bring it to the front. If there are no Do-file Editor windows open, it will create a new Do-file Editor window.

-------- update 04apr2012 -----------------------------------------------------

1. Online help and the search index have been brought up to date for Stata Journal 12(1).

2. alpha with options item and label and with the Results window sized greater than 99 characters wide reported incorrect observation totals. This has been fixed.

3. areg with factor-variable notation would, on rare occasions, fail to omit a collinear variable, resulting in a "matrix not positive definite" error. This has been fixed.

4. asclogit with option vce(cluster clustvar) sometimes did not converge when it should have. This has been fixed.

5. In Stata/MP, functions bofd() and dofb() could produce erroneous results or crash Stata. Passing the same arguments to one of these functions would yield correct results some of the time, yield incorrect results at other times, or cause Stata to crash, with essentially no ability to predict which of the three outcomes would occur at any particular function invocation. Stata/MP users who have used these functions should verify datasets and results that depend on proper conversion between business and daily calendar dates. These functions have been fixed.

6. import excel has the following fixes:

a. import excel imported an Excel column with only numeric cells and empty string cells as a string variable. This has been changed to import the column as a numeric variable.

b. import excel ignored Excel elapsed time formats. This has been fixed. Excel elapsed time formats are now imported as Stata datetime format %tc.

7. marginsplot working with more than one factor variable in margins results failed to identify an x-axis variable if it was a factor variable and the base was fvset. This has been fixed.

8. nbreg ignored option vce(opg) and reported the variance based on the observed information matrix instead of the outer product of the gradients. This has been fixed.

9. predict, score after glm, family(binomial) link(logit) caused an error if the dependent variable was missing in the observations that predict was used on. This has been fixed.

10. sem with option noivstart now allows the specified model to have more latent variables than observed endogenous variables. When option noivstart is not specified, sem will continue to report an error if there are more latent variables than observed endogenous variables.

11. tsappend, last() tsfmt() when the data had been tsset to have a panel variable carried over the last observed value to the appended observations rather than assigning missing values to them. This has been fixed.

12. ttest, by() with a string by-variable containing values with leading blanks issued a "more than 2 groups found" error. This has been fixed.

13. twoway area with option horizontal drew the plot incorrectly after the 30jan2012 update. This has been fixed.

14. use with varlist could cause Stata to crash if the command returned with an error. This has been fixed.

15. xtmixed, xtmelogit, and xtmepoisson failed to post the standard r() results associated with the coefficient table (ereturn display) when the fixed-effects table was reported. This has been fixed.

16. (Windows) In rare cases, browse or edit specified with if could cause Stata to crash when there were no data in memory. This has been fixed.

17. (Windows) Exporting an SEM graph to a bitmap format such as PNG resulted in a thin black line along the right and bottom edges of the graph. This has been fixed.

18. (Windows) In rare cases, update all could exit with error r(2101) even when the executable was named correctly. This has been fixed.

19. (Mac) When loading saved preferences while a Do-file Editor window was already open, the Do-file Editor window would get centered on the screen instead of using the saved position from the saved preferences. This has been fixed.

20. (Windows and Mac) Calling Stata in batch mode before entering a license and authorization key would cause Stata to hang in an infinite loop. This has been fixed.

-------- update 15mar2012 -----------------------------------------------------

1. The following commands now label standard errors as "Robust" instead of "Semirobust" when the vce(robust) option is used:

binreg, irls or glm, irls with canonical links xtgee with canonical links xtlogit, pa xtpoisson, pa xtreg, pa

Under the canonical link, "Semirobust" and "Robust" standard errors are equivalent.

2. biprobit has the following fixes:

a. biprobit with constraints reported the likelihood-ratio test of rho=0 rather than a Wald test. This has been fixed.

b. biprobit in the seemingly unrelated form produced an error when option iterate() was specified. This has been fixed.

3. bootstrap and bstat have new option ties.

4. collapse could produce incorrect summary statistics when stat was last(), first(), lastnm(), or firstnm() and the dataset had been sorted by more than one variable. This has been fixed.

5. contrast and pwcompare, after commands that have an implied intercept in the main regression equation, such as ologit and oprobit, incorrectly flagged some contrasts, pairwise comparisons, and margins as "(not estimable)". This has been fixed.

6. gmm has the following fixes:

a. gmm exited with an error message if option vce(cluster clustvar) was specified and the user had previously xtset the data with a panel variable but not a time variable. This has been fixed.

b. gmm exited with an error message if option xtinstruments() was used and 1) one or more panels was eliminated from the estimation sample because of missing data or 2) a panel had nonmissing data for only the observation corresponding to the last time period in the estimation sample. These issues have been fixed.

7. logistic with option log caused a syntax error, "option nolog not allowed". This has been fixed.

8. nlsur was documented as allowing the user to pass optional arguments to the function evaluator program. However, nlsur ignored these options and did not pass them to the function evaluator program. This has been fixed.

9. sem with option method(mlmv) sometimes produced a Mata trace log error instead of reporting that the starting values were not feasible. This has been fixed.

-------- update 06feb2012 -----------------------------------------------------

1. The 30jan2012 update to gmm introduced the following bugs, which have now been fixed.

a. gmm, twostep accepted iweights but did not handle them correctly under two-step estimation. This has been fixed.

b. gmm moment_prog exited with an error when multiple moment equations were used with analytic derivatives. This has been fixed.

2. lrtest issued a "samples differ" error when observations were added to the dataset between fitting the models that were being compared. The estimation samples do not actually differ when the additional observations are not used in fitting the second model. The error is no longer produced in this case.

3. The 30jan2012 update introduced a bug where user-written ado-files saved with legacy Mac end-of-line delimiters resulted in an unexpected end-of-file error when loaded. This has been fixed.

4. (Mac) Exporting a graph as a PDF using command form graph export ~/filename.pdf resulted in an error because the ~ was not converted to the user's home directory. Only the PDF format was affected. This has been fixed.

5. (Mac) Exporting a graph as a PDF with option name() always exported the topmost graph instead of the named graph. Only the PDF format was affected. This has been fixed.

6. (Linux) A library change in some of the latest releases of Linux distributions could cause Stata's memory to become corrupt. This behavior has been observed in CentOS 6.2 and Ubuntu 11.10 but not in any prior versions. The data corruption was isolated to variables that were loaded when a varlist was specified with command use, and even then was unlikely to occur. use without a varlist was not affected. This has been fixed.

-------- update 30jan2012 (Stata version 12.1) --------------------------------

1. Online help and the search index have been brought up to date for Stata Journal 11(4).

2. qreg has improved standard-error estimates. There are three new estimators available for estimating the standard errors: the fitted-value-based estimator, the residual-based estimator, and the kernel estimator. In addition, three new bandwidth choices are available.

The new default standard errors will provide better coverage than the old standard errors. Use version 12: qreg ... to reproduce the old standard-error estimates.

3. Normal random variates generated by rnormal() may be different between version 11.2 and 12.0. This has been fixed. The normal variates generated in version 12.1 will match those of 11.2. Use version 12: set seed # to reproduce normal random variates generated under version 12.0.

4. xtunitroot now accepts time-series operators.

5. assert has new option fast, which forces the command to exit when it encounters the first contradiction.

6. asmprobit and asroprobit with options intburn(#) and favor(speed) (the default) ignored the starting index specified in intburn(). This has been fixed.

7. business calendars have the following fixes:

a. An incorrect business calendar file -- a .stbcal file containing an error -- could cause Stata to loop endlessly when asked to display values formatted with that business format. This has been fixed. When a value is formatted with a format containing an error, the underlying numerical value is shown.

b. Some correct business calendars were mistakenly considered by Stata to be in error and thus produced error messages. This has been fixed.

8. clogit and xtlogit with option fe did not omit factor variables that have no within-group variance. This has been fixed.

9. collapse with statistic lastnm if by() was not specified sometimes selected a missing value instead of the last nonmissing value. This has been fixed.

10. estimates table has the following fixes:

a. estimates table when used with an interaction element specified in option drop() or keep() resulted in an invalid operator error even when the interaction term was specified correctly. This has been fixed.

b. estimates table when used with time-series operators in option drop() or keep() incorrectly reported that the coefficient did not occur in any of the specified models. This has been fixed.

c. estimates table with option equations() and results from sem would report an invalid operator error. This has been fixed.

11. glm incorrectly reported a syntax error if option vce(unbiased) was specified in addition to vce(robust) or vce(cluster clustvar). This has been fixed.

12. gmm now accepts iweights. iweights are treated like fweights except that iweights may be noninteger. When iweights are used, the reported sample size is the sum of the weights rounded down to the nearest integer.

Also, an error message is now issued if gmm encounters missing values when calculating the user-supplied analytic derivatives.

13. import excel and export excel have the following fixes and new features:

a. import excel and export excel now have new option locale(), which allows you to specify the locale of the Excel workbook.

b. import excel now has option case(). This option works with option firstrow and takes lower, upper, or preserve depending on what case you want your variable names to be.

c. import excel and export excel did not allow the documented command abbreviations. This has been fixed.

d. import excel did not always reset file handle tracking if an error occurred when reading the file. This has been fixed.

e. (Mac and Unix) import excel and export excel did not respect ~ in paths. This has been fixed.

14. margins after xtpoisson with option fe incorrectly reported margins and marginal effects of factor variables as "not estimable". This has been fixed.

15. marginsplot has the following fixes:

a. marginsplots of margins results using factor variables in option dydx() had unwanted base-level plots since the 13oct2011 update. This has been fixed.

b. marginsplot failed to identify the x-axis variable if it was a factor variable and the base was fvset. This has been fixed.

16. mata has the following fixes and new features:

a. The maximum number of numeric literals in a single Mata function has been increased from 100 to 2000.

b. Long sequences from Mata function ghalton() could deviate from the true Halton sequence. This has been fixed.

c. When there was an error, Mata function _st_addvar() returned with the Stata return code itself instead of the negative of the appropriate Stata return code as indicated by the documentation. This has been fixed.

d. (Windows) Mata function dir() returned file and directory names in lowercase. This has been changed to respect the original case.

17. mean, proportion, ratio, and total neglected to post the coefficient table results to r() when run quietly even when set coeftabresults was on. This has been fixed.

18. mi_estimate, post: issued a "last estimates not found" error when used with multilevel commands, such as xtmixed with more than two levels. This has been fixed.

19. ml has the following fixes:

a. ml with constraints used with option technique(dfp) or technique(bfgs) caused a conformability error in Mata. This has been fixed.

b. ml used with time-series operators but without option nopreserve resulted in the "initial values not feasible" error. This has been fixed.

20. After the 13oct2011 update, odbc load did not remember file information correctly for "MS Access Database" data source names. This has been fixed.

21. predict after xtmixed, xtmelogit, and xtmepoisson now reports a more informative error message when the predicted statistic requires random-effects estimation and the data used in estimation are not in memory.

22. _rmdcoll when supplied with a depvar that was constant and zero gave an unhelpful error message. This has been fixed.

23. rocreg, probit ml with pweights and multiple classification variables produced an error message. This has been fixed.

24. The F statistic reported by sdtest was not formatted properly when the decimal point was set to a comma instead of a period. This has been fixed.

25. sem has the following fixes and new features:

a. sem is now more convergent because of improvements in the computation of analytic derivatives.

b. sem now uses new algorithms that dramatically reduce the amount of memory required for models with many observed variables.

c. sem with a string variable specified in option vce(cluster ...) caused a Mata run-time error complaining that a real vector was required. This has been fixed.

d. sem specified with hierarchical latent structural paths and observed exogenous variables caused a Mata trace-log error. This has been fixed.

e. sem neglected to switch to noxconditional when constraints were applied to the mean of an observed exogenous variable. This has been fixed.

f. sem computed incorrect standard errors for the reported means and covariances of observed exogenous variables when sem remained xconditional. sem now switches to noxconditional when an observed exogenous variable is specified in option mean(), variance(), covariance(), or covstructure().

g. The SEM Builder could not fully edit variables whose labels contained quotes. Some edits worked, others did not, and selecting such variables produced error messages. This has been fixed.

26. If set segmentsize was used after set max_memory, Stata could fail to load any dataset with size larger than the new segmentsize. This has been fixed.

27. stsplit when used with after() and every() incorrectly split observations that occurred before the time specified in after(). This has been fixed.

28. twoway area now uses an internal drawing routine that is significantly faster over large datasets.

29. twoway contour drew an extra line at the bottom of the legend if the smallest cut was less than the minimum of the z value. This has been fixed.

30. use and merge will now exit with an error if the dataset on disk contains an invalid data type.

31. xtmelogit and xtmepoisson do not support weighted estimation. When fweights or pweights were specified, the weights were ignored and the unweighted results were reported. An appropriate error message is now issued.

32. A do-file containing ASCII character 255 would cause that do-file to terminate upon processing of that character. This has been fixed.

33. The varlist controls for dialogs sometimes were not displayed correctly when listing a large varlist. This has been fixed.

34. The Data Editor has the following fixes:

a. The 13Oct2011 update modified the behavior of pasting so that numbers with leading zeros preserved the leading zeros by importing the value as a string. However, the logic was flawed, causing a value of zero to also import as a string. This has been fixed.

b. Paste Special did not handle carriage returns embedded inside double quotes. Now carriage returns are processed as part of the string if the option "Double quotes bind tokens" is checked.

35. (Windows) A setting has been added to the Do-file Editor's Preference dialog that allows the behavior for .do file extensions to be changed so that the default is execute. Previously, the default behavior in Stata 12 was changed to always edit do-files.

36. (Windows) Typing characters in the Variables window now places the keyboard focus in the Command window, leaving the cursor at its previous location. Previously, the cursor was always moved to the end of any existing text. Note that the behavior for pressing the space bar has not changed; it still sends the selected variables to the Command window.

37. (Mac) Printing from the Results or Viewer windows now prevents partial lines from being printed regardless of paper size.

38. (Mac) Mac OS X 10.7 (Lion) introduced a feature called automatic text substitution in text controls that allows a user to type text such as tm and have it replaced by the trademark symbol. By default, it is turned on by the OS but it can interfere with Stata's ado-language. Automatic text substitution is now turned off in the Do-file Editor and the Command window.

39. (Mac) Mac OS X 10.7.2 (Lion) introduced a bug that causes the Zoom combobox in the SEM Builder's toolbar to crash Stata. This bug does not affect Stata running on earlier versions of Mac OS X. The appearance of the combobox has been changed to avoid this bug in Lion.

40. (Mac) The Break menu item and keyboard shortcut (Command+.) would always be disabled if the main Stata window's toolbar was hidden when Stata started up. This has been fixed.

41. (Mac) When saving an SEM path diagram, the default filename would have a file extension of .gph instead of .stsem if the Finder was set to always show file extensions or the Save dialog's Hide File Extensions checkbox was unchecked. This has been fixed.

42. (Mac) In very rare cases, Stata for Mac could crash if a browse or edit command was issued and resulted in an error (such as incorrectly typing a variable name). This has been fixed.

43. (Mac) The Variables Name field of the Data Editor's Properties pane was not editable unless a variable in the Data Editor's Variables pane was selected even though there was a single variable selected in the Data Editor's grid pane. This has been fixed.

-------- update 10nov2011 -----------------------------------------------------

1. gmm would on rare occasions exit with an error message when option xtinstruments() was specified and the minimum requested lag was greater than one. This has been fixed.

Also, we have tweaked the scaling bounds used by gmm's numerical derivative taker to better handle some difficult problems. Most users will notice no differences, but a few problems that failed to work previously should now work. Moreover, an error message is now issued if gmm encounters missing values when computing numerical derivatives.

2. ivregress incorrectly treated endogenous regressors that used time-series operators as exogenous when collinear variables and factor variables were specified as exogenous regressors (varlist1). This has been fixed.

3. marginsplot after the 13oct2011 update put option level()'s specification in the graph title instead of the specified confidence-level value. This has been fixed.

4. In styles mlong and flong, mi update (or any other mi command that runs mi update automatically) would unregister all passive variables whenever any passive variable was dropped using command drop. This could result in a loss of imputed data for some of the unregistered passive variables in style mlong. This problem did not occur if mi xeq: drop was used to drop a passive variable instead of drop. This has been fixed.

5. mi varying issued a "type mismatch" error in styles flong and flongsep when a string variable varied between imputations. This has been fixed.

6. mi xeq has the following fixes:

a. mi xeq now issues an error if collapse is attempted to be used with it.

b. mi xeq did not allow survey commands to be executed on individual imputations. This has been fixed.

7. rocreg failed when ctrlcov() was specified with bootstrap standard errors. This has been fixed.

8. sem with variances constrained to zero sometimes reported the "initial values not feasible" error, even when the model was properly specified. This has been fixed.

-------- update 13oct2011 -----------------------------------------------------

1. Online help and the search index have been brought up to date for Stata Journal 11(3).

2. help no longer displays a message box before doing a search if the help term was not a Stata command.

3. estat ginvariant after sem reported noninformative rows of dots for testing group invariance of latent variable means when the default constraints were used. This has been fixed to no longer show these rows.

4. The icd9 databases have been updated to use the latest version (V29) of the ICD-9 codes, which became effective on 1 October 2011.

5. import excel has the following fixes:

a. import excel imported cells with formatted text as empty strings. This has been fixed.

b. import excel imported text cells containing Unicode characters not supported by the current code page as empty strings. This has been fixed so that the Unicode characters are replaced as question marks.

c. import excel imported inline string and numeric cells in .xlsx files as empty string and missing values. This has been fixed.

6. margins after manova with option asbalanced with the default prediction was producing incorrect margin values when there was a factor variable in the model but not in the margins list. This was also the case if option at() was used to set the relative frequencies of a factor variable in the model. This has been fixed.

7. marginsplot has the following fixes and new features:

a. marginsplot ignored the level() and mcompare() information stored in r() after margins, yielding unadjusted confidence intervals using the default level setting. This has been fixed.

b. marginsplot now accepts options level() and mcompare(). These options affect the plotted confidence intervals.

8. odbc load now remembers your file information when using the "Excel Files" or "MS Access Database" data source names.

9. stcurve after streg with omitted variables produced missing values and blank graphs. This has been fixed.

10. The Data Editor has the following fixes:

a. Paste now respects set type.

b. Paste no longer treats numbers with leading zeros as numeric.

c. Paste did not create variable labels when the variable names were modified to form legal names in Stata. This has been fixed.

d. With the Data Editor open, any program or do-file that both renamed a variable and decreased the total number of variables could cause Stata to crash. This has been fixed.

e. (Mac and Unix) From the Data Editor's Paste Special dialog, the checkbox labeled 'treating sequential delimiters as one' could cause Clipboard data to be misinterpreted. This has been fixed.

11. (Windows) The dialog box for sktest caused Stata to crash when the Expression Builder was launched from the if/in tab. This has been fixed.

12. (Mac) Opening a do-file using the Open button in an existing Do-file Editor window would cause Stata to crash only when run on Mac OS X 10.5 (Leopard). This crash does not happen on Mac OS X 10.6 and Mac OS X 10.7 and does not happen if you use File > Open to open a do-file even on Mac OS X 10.5. We have changed the Do-file Editor to work around this issue with Mac OS X 10.5.

13. (Unix) When you click on a menu item in the Stata Help menu, a new Viewer is created. Before, it overwrote the current active Viewer.

-------- update 15sep2011 -----------------------------------------------------

1. ereturn display and some estimation commands would report omitted coefficients even when the showomitted setting was off or the noomitted option was specified. This has been fixed.

2. estimates table has improved parsing logic for options keep() and drop().

3. Estimation commands with empty cells in interaction terms of the model specification crashed Stata if the empty cells were dropped with set emptycells. This has been fixed.

4. graph export failed to set the font face when the output format was PDF. This has been fixed.

5. import excel and export excel could crash Stata if the Excel file contained a chart-only sheet. This has been fixed.

6. import excel with option allstring could import numeric values with less precision into string. This has been fixed.

7. import excel treated columns with numeric cells and empty cells as string columns instead of numeric columns. This has been fixed.

8. The maximum length of the Mata library path was increased from 256 to 4096 characters.

9. Prefix commands were ignoring misspecified if conditions when they should have been reporting an error. This has been fixed.

10. rename now updates tsset and xtset information in the data when the time or panel variable is renamed.

11. Zooming in and out from the SEM Builder now keeps the currently visible content centered in the window.

12. sem exited with a Mata conformability error when used with method(mlmv) and an estimation sample that had a missing-value pattern where all observed variables but one endogenous variable are missing. This has been fixed.

13. Single-equation estimation commands that allow options vce(jackknife) and vce(bootstrap) reported a syntax error when factor-variable or time-series operators that contained parentheses were used. For example,

. sysuse auto . regress mpg b(last).rep78

reported "variable b not found". This has been fixed.

14. ssd build exited with an error when the original data contained a value-label name that matched the name of the variable group(). This has been fixed.

15. stcox with option efron took a long time to compute when there were many tied failures. This has been fixed.

16. test, test after anova, and test after manova have new dialogs providing access to the equation names and coefficients of the last estimation.

17. translator query failed to display translator Graph2pdf's parameter settings. This has been fixed.

18. update all when called from batch mode would prompt the user before closing Stata. The behavior has been changed so that interactive prompts are now suppressed.

19. Adding variables to the Command window by clicking on a variable in the Variables window now prefixes a space character only when necessary.

20. In the Data Editor, in very rare circumstances, Paste Special could crash Stata. This has been fixed.

21. (Windows) window manage close viewer _all did not close any Viewer windows when it should have closed all of them. This has been fixed.

22. (Mac) Stata for Mac has the following new features and fixes:

a. Stata's file dialogs now support opening symbolic links to files or directories.

b. The Variables and Review windows now only pass a Return key press to the Command window when they have the keyboard focus.

c. The Do-file Editor will now open a do-file in the current tab or window if it is empty and unsaved rather than always create a new tab or window.

d. The gesture preferences are no longer displayed in the Preferences dialog when running under Lion because they are unsupported.

e. Launching the GUI version of Stata from a command line to execute a do-file caused the Break key to be disabled until the do-file had completed. This has been fixed.

f. In tabbed windows, the Close button for a tab is hidden until the mouse cursor is over a tab. Stata now always displays the Close button for the active tab unless it is the only tab in a window.

g. After updating ado-files or documentation only, Stata would not relaunch. This has been fixed.

23. (Unix) Dragging and dropping a long command from the Review window to the Command window crashed Stata for Unix. This has been fixed.

24. (Unix) The Do-file Editor did not support extended ASCII characters in do-files. This has been fixed.

-------- update 01sep2011 -----------------------------------------------------

1. The 01sep2011 Stata 11.2 update item 1 has now been applied to Stata 12. Other applicable items from that update were applied previously.

2. Prefix commands that produce replication-based variance estimates (such as bootstrap) when used with posted results from margins with option at() reported a "no observations" error after completing all the replications. This has been fixed.

3. Some estat commands after sem with method mlmv and xconditional results would error-out with a Mata trace log. This has been fixed.

4. In the SEM Builder, observed variables whose names begin with a capital letter were treated as latent when estimation was performed. This has been fixed.

5. In the SEM Builder, the property dialog boxes for variables and paths would not allow initial values to be removed. This has been fixed.

6. sem specified with models with three or more endogenous variables and some of them used to predict other endogenous variables would yield different standard errors or even fail to converge depending on the order in which the paths were specified. A typical difference in the standard errors was in the third decimal place and beyond, depending on the relative scales of the observed variables. This has been fixed.

-------- update 24aug2011 -----------------------------------------------------

1. blogit did not allow option vce(cluster clustvar). This has been fixed.

2. Business calendars with month or weekday names containing capital letters would fail to load. This has been fixed.

3. marginsplot now recognizes margins results in e() when r() no longer contains results from margins. margins will post its results to e() when option post is used.

4. Mata's optimize() with the Nelder-Mead technique and constraints would error-out with a Mata trace log. This has been fixed.

5. query now displays the setting coeftabresults as an output setting. This setting controls whether coefficient table results are stored in r().

6. roctab with option detail produced unaligned tables when the classification variable had long value labels. These labels are now abbreviated if needed, and a new option, nolabel, has been added to suppress displaying the value labels.

7. sspace, when time-series operators were combined with the e. error operator on the dependent variable in an obs_efeq equation, produced an error message. This has been fixed.

8. svy ivregress reported an invalid operator error when factor-variables notation was used in the list of instrumental variables (varlist_iv) but not in the list of exogenous variables (varlist1). This has been fixed.

9. When pasting data into the Data Editor, sequential delimiters were treated as one delimiter, causing missing cells to be omitted. This has been fixed.

10. (Windows) Updating Stata over a network share failed with error r(695), leaving the Stata installation unchanged. This has been fixed.

11. (Mac) When resizing the main Stata window, the Results view will now always be the main view that gets resized regardless of its location in the main Stata window.

12. (Mac) The keyboard shortcut for "Do to Bottom" for the Do-file Editor has been changed to Ctrl-Cmd-D because the previous keyboard shortcut conflicted with the system keyboard shortcut for hiding/showing the Dock.

13. (Mac) If tabbed windowing for the Do-file Editor is turned off and there are multiple Do-file Editor windows, Stata could be confused as to which Do-file Editor is active and perform menu and toolbar operations on the wrong window. This has been fixed.

14. (Mac) Translating SMCL to PDF has been reenabled.

15. (Mac) Translating SMCL to PS could cause Stata to crash. This has been fixed.

16. (Unix) Shift-clicking objects in the SEM Builder would not properly enable/disable items in the Object > Align menu. This has been fixed.

-------- update 08aug2011 -----------------------------------------------------

1. compress and commands that internally used compress could fail, causing future commands to issue a "variable not found" error. This has been fixed.

-------- update 05aug2011 -----------------------------------------------------

1. A serious memory management bug has been fixed. The bug was unlikely to occur. If it occurred, it usually caused a crash. It could, however, cause data to be corrupted. We had only four reports, all involving crashes. If a dataset was below 32 MB on 64-bit computers or 16 MB on 32-bit computers, the bug could not occur.

2. stcox using options efron and vce(robust) could take a long time to compute for large datasets with many tied failure times. This has been fixed.

3. (Mac) Changing the results color for the Results window had no effect. This has been fixed.

4. (Mac) The keyboard shortcuts for Shift Left and Shift Right in the Do-file Editor's contextual menu did not match the main menu's keyboard shortcuts (which take precedence). This has been fixed.

5. (Mac) Graphs drawn using polygons would render with thick lines when exported to PDF or a bitmap format. This has been fixed.

6. (Windows) Some tooltips for copying to the Clipboard could incorrectly show up as "Copy Diagram". This has been fixed.

7. (Windows) On some international keyboards, characters that required the right-side Alt key to be pressed, such as { and } on German keyboards, could not be typed in the Command window or the Do-file Editor. This has been fixed.

8. (64-bit Windows) query memory displayed an extra letter "c" when reporting memory sizes. This has been fixed.

-------- update 26jul2011 -----------------------------------------------------

1. Online help and the search index have been brought up to date for Stata Journal 11(2).

2. The 19jul2011 Stata 11.2 update items 2, 3, 4, 6, 7, 11, 12, 13, 14, 15, 17, and 21 have now been applied to Stata 12. The other items from that update were applied to Stata 12 prior to its initial release.

3. margins with options vce(unconditional) and noestimcheck reported a "varlist required" error if a factor variable was also specified in one of the marginal effects options (dydx(), dyex(), eydx(), or eyex()). This has been fixed.

4. (Stata/MP) mlmatbysum produced wrong results if the global macro $ML_w was empty. This has been fixed.

5. notes search issued an r(111) error if the search string in question was not found in a note. This has been fixed.

6. The new SEM Builder has enhancements and bug fixes, such as the following: settings affecting appearance are now remembered between Stata sessions (enhancement), and paths connecting error variables now join better with the containing circle (bug fix).

The thirty-two enhancements/fixes are not enumerated here because they were made on the date of release.

7. sem has the following fixes:

a. sem would sometimes report structural equations out of the order in which they were specified. This has been fixed.

b. sem with a svy prefix or with options vce(bootstrap) or vce(jackknife) failed to report the standardized model parameters when option standardized was specified. This has been fixed.

8. (Mac) In a tabbed Viewer window, dragging the last tab out of the window now destroys the empty window. Previously, the empty window wasn't destroyed, which could cause problems when a new Viewer was created.

9. (Mac) The Do-file Editor automatically saves backups of unsaved do-files. If Stata crashes, then the next time Stata is launched, the backup is automatically opened in the Do-file Editor. Previously, if the backup was closed without saving changes, the backup would be opened every time Stata was launched. This has been fixed; Stata now removes the backup file.

10. (Unix) The Stata GUI issued GTK warnings if your dataset contained extended ASCII characters. This has been fixed.

-------- previous updates -----------------------------------------------------

See whatsnew11to12.

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


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index