st: Discrete-Time Event History Analysis (Logit) with RE

[Andreas Schiffelholz <Andreas.Schiffelholz@gmx.de>]

Hello Statalist community,

I'm currently calculating a discrete-time event history analysis, using 
a logistic regression with duration dummies. The subject of analysis is 
an unbalanced panel of ~1000 company years (t: 9 years, x: 160 
companies). I'm analyzing the influence factors of strategic changes 
(dep. var. in year x: 0 = no change in strategy in year x; 1 = change in 
strategy in year x). To account for duration effects I include Dummies 
in the model indicating the length of the current strategy phase ("1 
year" as omitted base category, "2 years", "3 years", "4 years", "5+ 
years") if the length of the strategy phase is known and another dummy 
variable if the length is not known.

Because there is more than one strategy phase per company in the sample, 
I want to calculate a random effects model in addition to the cluster 
robust pooled logit model, with the company ID as the second level. If I 
calculate the random effects model (xtlogit, re) including only the 
duration dummies, there is hardly any difference to the corresponding 
pooled model with the "Likelihood-ratio test of rho=0" being far from 
significant and if I add more variables such as industry dummies, fiscal 
year dummies or company age there is no difference to the pooled 
coefficients at all, with :

Likelihood-ratio test of rho=0: chibar2(01) =     0.00 Prob >= chibar2 = 
1.000


Does this just mean, that the random effects model is not needed and the 
pooled cluster robust model is suitable for my analysis or is it 
indicating a severe error in the specification of the model?


Thanks a lot!
Andreas



*
*   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/