Re: st: Multiple Imputation in Longitudinal Multilevel Model

["JVerkuilen (Gmail)" <jvverkuilen@gmail.com>]

On Wed, Mar 6, 2013 at 12:47 PM, Anthony Fulginiti
<fulginitipsy@yahoo.com> wrote:
 In any case, I simplified the model for the post but had followed
your suggestions and used a host of variables (known to relate to the
outcome variable in the literature) in the imputation model and the
xtmixed model.
>

You may not necessarily want them in the -xtmixed- model.


>
> 1)I will increase the number of imputations.

For the purposes of model building and checking the imputation model
it's not crazy to work with a smaller number. Once you're satisfied
with the final imputation model then it's worth boosting that number
up.


>
> 2)I did not include time in my imputation model because my understanding was to perform the imputation in wide format rather than the long format (when reshaping from long to wide, the time indicator variable is dropped until I subsequently reshape for the use of the mim command). Perhaps if I follow your suggestions and use the MI feature, I can incorporate time and the quadratic for time since I found that polynomial function is the best fit for the existent data.
>

Usually it's a good idea to make sure that important structure such as
time and clusters are represented, but often leave a little extra room
for wiggle. Wide tends to destroy those trends, though. Um... Gary
King and one of his collaborators wrote something on this. I believe
it's on the Amelia web page:

     http://gking.harvard.edu/files/abs/pr-abs.shtml

Honaker, James, and King, Gary. 2010. What to do About Missing Values
in Time Series Cross-Section Data. American Journal of Political
Science 54, no. 3: 561-581.




> *A fundamental issue that I was trying to address is that I know that certain variables, such as symptoms relate to the outcome.  However, I have missing values in several of those relevant independent variables as well as in the outcome variable. My understanding was that the chained equation approach (with ice and then the mim command) would work well when missing values for both independent and dependent variables.

I may be misunderstanding your question but there are true independent
variables that shouldn't be missing, but each dependent variable gets
its chance in the sun to predict the others.



> I will further review the MI manual but let me know if you have any additional thoughts or helpful hints on the matter. I appreciate the insight and perspective on such challenging material.

Yes, these models are tough.

-- 
JVVerkuilen, PhD
jvverkuilen@gmail.com

"It is like a finger pointing away to the moon. Do not concentrate on
the finger or you will miss all that heavenly glory." --Bruce Lee,
Enter the Dragon (1973)

*
*   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/