st: Pearson's goodness of fit test for logistic regression using complex survey data

[Salazar Altamirano Saidé Aranzazú <ssalazar@cnbv.gob.mx>]

Hi Statalist users,

Hoping that you can offer some advice about the following issue.

I would like to apply the Pearson's goodness of fit test after a logistic regression. I am aware the test is valid under the assumption of independent and identically distributed observations. However, my data come from a complex sampling design. I know Hosmer and Lemeshow developed an alternative to  their standard goodness of fit test, taking into account the survey sampling weights and design (svylogitgof or estat gof in recent Stata versions). I would like to know if there is a similar alternative to estimate the standard Pearson statistic after a logistic regression using complex survey data. My interest in obtaining the Pearson statistic (i.e. a "modified" version of it for complex survey data) comes from the fact that my model has a small number of covariate patterns. As I understand, one of the assumptions for the Hosmer-Lemeshow test be valid, is to have a number of covariate patterns almost equal to the number of observations. Actually, I have around 1400 observati!
 ons (representing a population of 66,000 under probability weights) with 300 covariate patterns (the independent variables are all dummies).

Thank you very much for all your help!

Saide
*
*   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/