Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Postgr3,interactions in models with squared terms


From   "G. Vazquez" <academic.demographer@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: Postgr3,interactions in models with squared terms
Date   Thu, 24 Sep 2009 10:45:18 -0700

Hi Statalist,
I am using postgr3 (Stata 10.1) and am unsure about two things: 1.
When using svy logistic, does postgr3 give predicted odds, and
predicted logits when specified with predict (xb)?, and 2. When a
square-term is included in the model, should it theoretically  be
allowed to vary using (asis)?
Here is an example:

Dependent: User (1) vs. non-user (0)

Independent: Nativity (3 categories), continuous age, several
dichotomous controls, and age-squared

The interaction, which is significant, is between nativity and age (continuous).

Theoretically, it makes sense to me to have the predicted odds of Y,
holding all continuous covariates constant at 0, except age-squared.

The syntax (?):
xi3: svy:logistic treat_mind i.nativity_yrs*age_p  i.race i.female
i.married agesq_r
postgr3 age_p, by(nativity_yrs) x(_Irace1=0 _Irace2=0 female=0
married=0) asis(agesq_r)

However, when I allow age-squared to vary, the slopes are no longer
different for the three groups, and the direction of the slopes
changes.

I have read up on postgr3 on the help and UCLA pages, but am still
unclear about the above issues and want to make sure that I am
understanding it correctly.

Thank you in advance for entertaining my question.

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index