Re: st: Subgroup mean imputation

[Nick Cox <njcoxstata@gmail.com>]

Given

egen classmean = mean(score), by(CODICE_CLASSE)

then (1)

clonevar score2 = score
replace score2 = classmean if missing(score)

or (2)

gen score2 = cond(missing(score), classmean, score)

It's good practice to keep your original variable unchanged. In (2)
use -double- or -long- if -score- is -double- or -long-.

Nick
njcoxstata@gmail.com

Barbara Romano

> Thanks Fernando,
> unfortunately it’s not working. I’m stuck with my missing .
> I never used looping, but do you think a loop could be a solution?
> Something that says stata for each missing in the variable classmean impute the value of classmean of another observation with the same classroom_id…

Fernando Rios Avila

>> What I usually do is to add the following command to what you already have:
>> egen classmean_imputed = max(classmean), by(CODICE_CLASSE)

Barbara Romano

>>> I’d like to contrast the results I get the mi imputation and the old fashioned mean imputation.
>>> I’ve a huge dataset of student test scores with a variable for classroom id.
>>>
>>> I used the egen command to create the subgroup mean for each classroom
>>> egen classmean = mean(score), by(CODICE_CLASSE)
>>>
>>> this create for each student with the same classroom code, BUT for the the students with a missing value for the score.
>>> I’d like to impute that classroom subgroup mean for the “missing students”.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/