🚀 go-pugleaf

hi all,

I need to estimate a hazard model using a Weibull distribution that
also has time-varying covariates.  In this case, at least one
covariate changes for each time period, so in a sense noone has
a spell longer than 1.  I've looked at the BMDP, LIMDEP and
SAS (both PHREG and LIFEREG) procedures and none of them seem
able to handle time-varying covariates AND a Weibull distribution.

I have found an article by Trond Peterson (Soc. Methods and Research,
1986) which says such models are possible and gives BMDP/FORTRAN
code for estimating an exponential model.  I could derive similar
code for the Weibull, but I would much rather have a tested,
easy procedure.

So, any advice?  Is there some way to either trick a proportional
hazards procedure or a Weibull-capable procedure into doing
this type of model.  Is there a package which handles these
directly?

thanks for your help,

Walter Davis <WDAVIS@UNCVM1>
Department of Sociology
UNC - Chapel Hill