Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: question about multiple imputation for repeated measures in panel data


From   Michael Chavez Reilly <michael.chavez.reilly@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: question about multiple imputation for repeated measures in panel data
Date   Tue, 4 Dec 2012 15:07:33 -0500

Greetings statalist gurus!

I’m a doctoral student working on my dissertation and I have a
question about the utility of multiple imputation for panel data with
repeated measures. I am fairly familiar with the multiple imputation
process as I have multiply imputed cross-sectional data using stata in
the past. My quandary is this -  I’m using panel data with 4 waves
(1993, 1994 1997 and 2003). Though my survey panel weights handles
respondent drop-outs, there are still between 10-25% of cases with
item non-response on some key predictors and outcomes. So, I figured
multiple imputation made sense.

I began constructing an iterated chained equation model using the mi
impute chained command in stata 12, but even after aggregating all my
dummy variables into comprehensive categorical variables to avoid
collinearity within the ICE model (and relegating other highly
correlated variables to auxiliary status)l, I still face the problem
of perfect prediction in repeated measures of variables (such as type
of job in 1994 and type of job in 1997, marital status in 1994 and
marital status in 1997, etc).

Stata tells me I can use the “augment” option (mauglogit, etc) as a
workaround, but given that I have a fair number of repeated measures
across the 3 to 4 waves I wish to impute, I’m wondering if
statistically this becomes problematic. Basically, mauglogit may end
up being used a lot in my ICE models as there are a number of repeated
measures across 3-4 waves of panel data.

Any thoughts, advice, suggestions or ideas on whether using MI with so
many repeated measures is legitimate would be incredibly appreciated!
For example, should I drop my quest and try Max likelihood estimation
inn MPLUS instead?

Thank you in advance for any help you can provide.

Best,
Michael

*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index