I guess that Joe has in mind here his -unblock- program published in STB-7.
I can't find a record of any sibling -block-.
As earlier posted, combining -reshape- and -expand- gives an official Stata
solution.
Nick
[email protected]
=====================================
Nick:
Well, you may be correct. However, but it seems that I also wrote programs
called -fold- and -unfold- as well. Those were the ones I remembered
when reading the email. Here is the -unfold- ado file, and its associated
help file. I notice now that it is written in version 3.0 code. Presumably
it
should work, but it may need updating. The logic remains the same so that
anyone wanting to amend it to current version code may do so. Of course,
one may simply use a combo of Stata's expand command. But I found using
this program (or block/unblock - they were published together) much easier.
I
have to admit that the program appears rather primitive -- it's 13 years old
now.
Joe Hilbe
*! version 3.0.0 4/23/92 Hilbe
capture program drop unfold
program define unfold
version 3.0
local varlist "req ex min(2)"
parse "`*'"
parse "`varlist'", parse(" ")
confirm new var _`1'
confirm new var _`2'
local k = _N
local k1 = _N + 1
quietly expand =2, clear
capture {
gen byte _`1'=0 in 1/`k'
replace _`1'=1 in `k1'/l
gen _`2'=`2'-`1' in 1/`k'
replace _`2'=`1' in `k1'/l
mac shift
mac shift
}
noisily di in bl " Blocked data unfolded..."
local rc = _rc
error `rc'
end
=========================================
================================================
Help of ^unfold^ : Program to unfold or unblock data sets
--------------------------------------------------------
Joseph Hilbe, Editor STB
The syntax for ^unfold^ is:
^unfold^ pos_var pop_var [rhsvar]
pos_var is a variable containing the number of positive responses for a
group or block. pop_var is the total cases in a block or group.
Right-hand-side covariates will be appropriately handled; you do not
need to name them on the command line.
Consider the following grouped data set:
infec cases cd4 cd8
1. 2 2 1 2
2. 4 7 0 0
3. 1 3 2 2
4. 0 2 2 0
The first observation tells us that there are two cases having the values of
1 and 2 for cd4 and cd8 respectively. Moreover, both of these
cases have a positive response (i.e. infection occurs). The second
observation represents a group consisting of four positive responses
out of seven cases for a covariate pattern of 0 0.
^unfold^ literally unfolds or ungroups a data set. In order to do this
two new variables are created.
^. unfold infec cases
Blocked data set unfolded...
^.l
infec cases cd4 cd8 _infec _cases
1. 2 2 1 2 0 0
2. 4 7 0 0 0 3
3. 1 3 2 2 0 2
4. 0 2 2 0 0 2
5. 2 2 1 2 1 2
6. 4 7 0 0 1 4
7. 1 3 2 2 1 1
8. 0 2 2 0 1 0
Notice that the first observation, having a covariate pattern of 1 2, shows
no cases with a non-positive outcome. Observation 5, having the same pattern,
records two positive responses (infections). This accords with what we saw
in the grouped format. Also notice that the newly created variables of
interest have the same name as the original grouped variables, except that
they have a leading underscore. Although the original variable could have
been discarded in the transformation, I decided to retain them in the
ungrouped data set in case a user may find them valuable. However, they are
not needed for most subsequent analysis.
Remember, when performing subsequent analysis, ^_cases^ is a frequency
weight. Hence, a logistic regression of the above data set would look like:
^logistic^ _infec cd4 cd8 [fw=_cases]
Of course, you may rename the underscored variables to anything you
like. Refer to ^sqv4^ for use with blocked logistic regression.
========================================================================
Joe Hilbe
> I once wrote an ado file for the STB called -unfold-, which went
together
> with -fold-. It is in -sqv4 in one of the old STB's. It should be easy to
find
> looking at the early STBs on Stata's web site. The -unfold- program was
> designed for this exact situation. The help file that goes with the
program
> advises how to use it appropriately. If you have difficulties locating
it, let me
> know privately and I'll send it to you.
>
> I also recall that there are now other Stata programs in existence that
can
> do the same thing, but cannot think of their names.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
