st: Testparm Command in Logistic Regression Analysis

[<isabel.lechner@vetsuisse.unibe.ch>]

Dear all, 

I'm performing a logistic regression with 5 categorical predictor variables. 

xi: logit status2 i.season2 i.sex2 i.age2 i.unit2 i.distance5

I intended use the- testparm- command  for each variable to get an overall significance for the variables in the model.

testparm i.season2
testparm i.sex2

Both of these commands are not working, as Stata tells me "no such variables".

But as soon as I change the baseline categories, the testparm command  works:
xi: logit status2 ib1.season2 ib1.sex2 ib1.age2 ib1.unit2 ib1.distance5
If I  then run the -testparm- command after, it gives me an output!

Any explanations for this? Did I miss something essential? My intention was to decide whether a variables stays in the model (backward selection procedure) based on the p-value of the testparm command. 
First  I tried to compare nested models with the - lrtest- command, but this didn't work out as I have different numbers of observations within the models!

Thanks for any help and clarifying!

Isabel


----------------------------------------------------------------------------------------------------------------


Isabel Lechner, Dr.med.vet.
Resident ECVPH

Veterinary Public Health Institute
Vetsuisse Faculty, University of Bern
Schwarzenburgstrasse 155
3097 Liebefeld, Switzerland

Tel. +41 (0)31 631 5733
Fax +41 (0)31 631 5749
isabel.lechner@vetsuisse.unibe.ch 
www.vphi.ch


*
*   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/