As I understand your slowly emerging problem: Not so.
A side-effect of mapping all but one value to missing will be to exclude all the changed observations from any subsequent model fit. Thus you end up using just one observation from each school, which you do not want at all.
The source you are using is not drawing on any information about how Stata works. What you are asking in effect is that Stata impute the missings on the fly within schools, but it won't do that.
Your original data structure is, again if I understand you, not only fine for your purpose but the only one that will be appropriate for your problem.
Otherwise put: everyone who answered gave good advice for what they understood to be your problem. But everyone was addressing problems that differ from your need.
Nick
[email protected]
Russel Wildeman
Because of the nested data structure, the value of the school-level
variables (ie those that have been aggregated) will be duplicated for the
same school (because for the same school, they must be the same), whereas I
wanted only one record (I must not say observation and this is what was
perhaps confusing-my apologies) per school. This should do the trick
I want to model class size and educational outcomes jointly and the source I
am using suggests it is computationally more efficient to have only one
record per school. I want to use a multilevel bivariate response model for
this task.
Martin Weiss
Actually, I had the same idea as Jeph initially, because Rusell requested
that "... one of the school-level variables (classize) ... have only one
observation per school", but I dismissed it as unlikely. If this is indeed
what Rusell wants, I would be interested to hear what kind of analysis
requires that the variable have only one non-missing value per unit (Schools
in his case).
Furthermore, he could then look at -egen, tag()- which would give him an
indicator variable which he could condition on in subsequent analysis
instead of -drop-ping the remaining cases outright.
Jeph Herrin
If I understand correctly,
bys school : replace classize=. if _n>1
Russel Wildeman wrote:
> Thank you Martin
>
> The data look like this
>
> School Student Classize female
> 1 1 35 1
> 1 2 35 0
> 1 3 35 1
> 1 4 35 1
> 2 1 23 0
> 2 2 23 1
> 2 3 23 1
> 3 1 39 0
> 3 2 39 1
>
>
> What I would like is for one of the school-level variables (classize) to
> have only one observation per school. Hope I am a bit clearer.
>
> Many thanks,
> Russell
>
>
>
>
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of Martin Weiss
> Sent: 16 December 2009 04:03 PM
> To: [email protected]
> Subject: st: AW: Creating single values from a nested data file
>
>
> <>
>
> Could be a case for -reshape- or -collapse-. For further help, you should
> post a meaningful excerpt of your data.
>
>
>
> HTH
> Martin
>
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected]
> [mailto:[email protected]] Im Auftrag von Marcello
Pagano
> Gesendet: Mittwoch, 16. Dezember 2009 15:04
> An: [email protected]
> Betreff: st: Creating single values from a nested data file
>
> For Russell:
>
> *From:* Russel Wildeman [mailto:[email protected]]
> *Sent:* 15 December 2009 07:36 PM
> *To:* '[email protected]'
> *Subject:* Creating single values from a nested data file
>
> Dear Listers,
>
> I am new to Stata and have the following question.
>
> I have education survey data where the data are ordered by school and
> student and more pertinently the variable "school." The data are ordered
> by the variable "school" and would give 0 0 0 0 1 1 1 1 etc. I have
> created appropriate school-level variables but the problem is that the
> values of the school-level variables are repeated for the same school,
> whereas I want only one record per school for some of the school-level
> variables. I tried various merging file schemes but none work. Any help
> is highly appreciated
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
