Dear Statalisters,
I'm estimation a Cox model for the 'hazard' to be a top performing academic=
researcher (in terms of publications). In my standard Cox model I have fou=
nd a strong dependence of current top performance on previous top performan=
ce. Before deciding that the academic research process is highly 'accumulat=
ive' in nature (earlier success leads to future success), I want to estimat=
e a frailty model to determine whether there is underlying individual heter=
ogeneity which would offer an alternative explanation (some people are just=
inherently more productive than others).
I'm running into a problem which I think is due to a limitation of Stata v8=
=2E0 Intercooled.
An example command:
=2E stcox male age age2 r_doc1 r_hdoc1 r_hl1 r_rest1 yrsrank headuni1 fullt=
iun tenure zapaft92 yrszap tchload pasttop2y, shared(persnr)
matsize too small
=A0=A0=A0 You must set matsize to at least 946, where 946 is the number of =
groups plus the number of covariates; see help matsize. r(908);
I have 1040 researchers in my sample and my interpretation of the above err=
or is that because I try to estimate a frailty parameter for each individua=
l, I'm exceeding Stata Intercooled's limitation of matsize=3D800. Estimatio=
n on a subsample works, so I assume this is the reason.
Can anyone confirm this or am I misunderstanding the shared frailty model? =
Any other ways out of this besides upgrading to Stata SE with a matsize=3D1=
1,000 limit?
Sincerely,
Stijn Kelchtermans
Doctoral Researcher
Katholieke Universiteit Leuven, Belgium
[email protected]
=A0
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
*
* 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/
