Re: st: weights and log likelihood

[Steve Samuels <sjsamuels@gmail.com>]

The log pseudo-likelihood value itself has no real bearing on survey
inference. You can't compare models by comparing the difference in log
likelihoods, for example. The contributions of each individual are
weighted by the probability weight, so that the log-likelihood total
estimates the one you'd get if you had data on every individual in the
population. Thus the big number isn't surprising.

I know little about -dprobit-, except that it is much less
capable than margins. If you are analyzing a multistage survey, then I
assume that you've -svyset- your data to get proper standard errors.

Steve

On Nov 14, 2012, at 2:12 AM, Elin Vimefall wrote:

Hi

I run a probit model using pweight.
However when the weights are introduced the Log pseudolikelihood becomes really large (-11413870).
Can some one help me  understand how the weights influence the Log pseudolikelihood ?

(If I instead run the dprobit, since I'm interested in the marginal effects, the Log pseudolikelihood becomes "normal" again)


Thankfull for all help i can get!

//Elin Vimefall
*
*   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/



*
*   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/