Dear S+ users:
I am fitting parametric models to a censored
data. Currently there are two choices in Splus - survReg or censorReg. My
understanding is that, based on the manual, the only difference is that censorReg is more powerful in handling complicated cases
such as left truncation/interval censoring. So, they should have the same performance
for a “simpler” data. Below I fitted an exponential model to a
right censored data, and there is a rather big inconsistence, especially the
intercept.
My questions are: 1) is there any trick in model
parameterization for these two functions? 2) if not,
which should I believe? 3) if yes, where should I get
more details?
Many thanks for your comments
Feng
censorReg(formula
= censor(t2, c) ~ age + vcd + iop
+ psd + cct, data = ""
style='mso-spacerun:yes'> distribution = "exponential")
(Intercept) age vcd iop psd cct
5.284502 -0.583164 -0.2693929
-0.1353391 -0.1331131 -1.022518
Log-likelihood: -688.2034
survReg(formula
= Surv(t2, c) ~ age + vcd +
iop + psd + cct, data = "" dist = "exponential")
Coefficients:
(Intercept)
age vcd iop
psd cct
1.573442 -0.2785216
-0.2576163 -0.1005846 -0.09998574 -0.6032965
Loglik(model)=
-493.4 Loglik(intercept
only)= -537
Feng Gao, Ph.D.
Division of Biostatistics, Washington University School of Medicine
Campus Box 8067, 660 S. Euclid Av.
St. Louis, MO 63110-1093
Phone: 314-362-3682
Fax: 314-362-3728
The materials contained in this e-mail are private and confidential and are the
property of the sender. If you are not the intended recipient, be advised
that any unauthorized use, disclosure, copying, distribution, or the taking of
any action in reliance on the contents of this information is strictly
prohibited. If you have received this e-mail transmission in error,
please immediately notify the sender.