Dear Peter,
Thanks for your comment.
If I include both the dummy variables as defined in the previous mail, they get dropped due to multicollinearity. I tried creating another set of dummy variables as follows:
hvptb: 1 if ptb, 0 if anything else
hvinf: 1 if infection, 0 if anything
normal: 1 if normal, 0 if all else
WIth these dummy variables in the regression with robust standard errors model, the interactions with time are not significant ( the main effects are still significant).
However in the gllamm model ( family(gamma), log link, i(id) ), the main effects and interactions are not significant.
Hence, I tried fitting 2 models: one comparing just the normal and the ptb and the second comparing the normal and infection.
Thanks again,
Leny
Leny Mathew
Data Analyst
Division of Perinatal Research
Department of Obstetrics/Gynecology
Drexel University College of Medicine
245 N. 15th Street, 17th Floor
Mail Stop 495
Philadelphia, PA 19102
Phone: 215-762-2069
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
