See also
SJ-7-2 dm0030 . . . . . . . . . . Stata tip 44: Get a handle on your sample
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Jann
Q2/07 SJ 7(2):266--267 (no commands)
tip on how to keep track of the observations that make
up your estimation sample
Nick
[email protected]
Eva Poen
I should add that -browse if e(sample)- works just as well, as long as
you haven't fitted another model in the meantime. e(sample) (like all
e() results) always refers to the most recently fitted model.
Therefore, if you want to compare which observations were used in two
different models, you have to generate an indicator variable along the
lines of -gen inmodel = e(sample)- before fitting the second model.
Eva Poen
> After estimation:
> gen inmodel = e(sample)
> browse if inmodel
>
> The function e(sample) returns 1 if the observation was used and 0 otherwise.
Jiang, Frank <[email protected]>:
>> I am having difficulty identifying observations used in a logit model.
>> The model includes several sets of dummy variables. Some observations
>> were automatically dropped because some dummy variables perfectly
>> predict failures, i.e. DV==0. As a result, the number of obs used by
>> the logit model is smaller than the number of obs in the dataset.
>> However, I don't know which obs were excluded. Is there a way to
>> identify the obs that were actually used in the model?
*
* 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/
