RE: st: MI IMPUTE MVN

["Jocelyn Crowley" <jocelync@rutgers.edu>]

Dear Stas:-

Thank you very much for your help.  I appreciate your taking the time to
answer my questions.

My data set is small, but my problem is missing values.  I have about 441
observations from a random sample survey with (data includes weights), but
casewise deletion results in only 340 cases.  I am using the data to
generate 4 logit models.  Some of my missing data are dummy variables.  I
would like to 1) generate summary statistics for my data (means, ranges,
standard deviations), 2) generate 4 logit models, 3) calculate predicted
probabilities given certain values of the independent variables. I just
don't know how to handle the missing value problem, and am open to
suggestions.

I would like to know how much you charge for consulting, and how you might
be able to help me.  I do not have grant money now, but would like to know
what your rate is anyway. 

Thanks!

Jocelyn





-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Stas Kolenikov
Sent: Friday, October 29, 2010 3:11 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: MI IMPUTE MVN

On Fri, Oct 29, 2010 at 2:14 PM, Jocelyn Crowley <jocelync@rutgers.edu>
wrote:
> 1.  If I will be using weights with my data when I run mi estimate
logistic
> commands, do I need to direct Stata to consider these weights when I first
> run mi impute mvn?  If so, how?

If it's not in the syntax diagram, then you cannot do this. I think
-ice- suite by Patrick Royston supports weights. If you have complex
survey data, then MI gives wrong answers, anyway. You'd need to go
into considerable lengths to make that right.

> 2.  I would like to calculate means and standard deviations for my imputed
> data.  mi estimate means calculates the means, but not the standard
> deviations of the variables.  How do I do this manually?

You would generate the squares of your variables. I think there's a
debate as to whether they need to be passive or imputed variables.
Either the MI manual or Royston's papers address this issue.

> 3.  Once I obtain my mi estimate logistic model, how can I manually
> calculate predicted probabilities given certain values of the independent
> variables?  Do I need to do this by hand, given the model's coefficients?

You would probably need to generate a set of predicted probabilities
for every set of imputed values, and then aggregate them. That's a
working solution; I cannot say whether it is a conceptually valid
solution.

-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.

*
*   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/


*
*   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/