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]

Re: st: mi impute: VCE is not positive definite


From   "JVerkuilen (Gmail)" <jvverkuilen@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: mi impute: VCE is not positive definite
Date   Tue, 30 Oct 2012 09:07:34 -0400

On Tue, Oct 30, 2012 at 1:53 AM, chong shiauyun <shiauyun416@hotmail.com> wrote:
> Hello all,
> I was using -mi impute chained- to impute my dataset. I have trouble with the variable citwlc_i, which is an interaction of cit*wlc. (cit is my temperament categories(3) and wlc is parentalal warmth latent class). I received the following error message:
>
>
> mi impute: VCE is not positive definite
>     The posterior distribution from which mi impute drew the imputations for citwlc_i is not proper when the VCE estimated from the observed data is
>     not positive definite.  This may happen, for example, when the number of parameters exceeds the number of observations. Choose an alternate
>     imputation model.
>
> I am not sure why this happens and what should I do to make the model works.Please advice.
>

As I said last week, you're hoping for this to work. It won't. The
error message you got is essentially saying that you have
collinearity.

How many N do you have? I guess that you need about 10-20 cases per
variable and I'd say you have a lot of variables. This is a ballpark
but if you go much lower than that you will run into trouble. Cut WAY
back on the variables. You have a ton of imputed variables here and
it's a testament to the MI program that it copes at all.

Cut out all the interactions for now. Try imputing with just main
effects. Add the interactions back after you've gotten a reasonable MI
run with main effects only.

Also you can almost certainly do with out the truncreg all over the
place. You can always put it back on if it's absolutely necessary but
it (a) slows things down due to needing iteration and (b) imposes more
assumptions on the structure. Again, add this back later.
*
*   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