The syntax you gave works for constraining kappa in the example below,
except for some reason when fitting a constant-only model.
In this example, when kappa is constrained to be 0 the starting values
for the estimation are not feasible. The log-normal model is the same as
the generalized gamma with kappa=0, so you can fit both of those to
perform a likelihood ratio test of kappa=0.
The options -anc()- and -anc1()- fit the ancillary parameters (or the
log of the parameter) as linear combinations of the variables specified,
which is not what you want here.
Jamie Griffin.
sysuse cancer,clear
stset studytime, f(died)
constraint 1 [kappa]_cons=0
streg drug, dist(gam) constraints(1)
constraint 2 [kappa]_cons=0.00001
streg drug, dist(gam) constraints(2)
streg drug, dist(logn)
streg drug, dist(gam)
streg, dist(logn)
streg, dist(gam)
constraint 3 [kappa]_cons=0.00001
streg , dist(gam) constraints(3)
>>> [email protected] 09/14/05 10:48 am >>>
Hi, listers
I'm trying to perform an lrtest on survival time data. Thus I need to
create a restricted model, which I would do using the -constraint-
command and adding the parameter constraints(#) to the -streg-
commmand.
However, this does not seem to work very well and stata still
calculates
the kappa-parameter. Please see attached printout.
I tried doing a workaround by specifying anc(kappa1) where kappa1 is a
variable containing only 1's, but computing that regression has my
computer running on it's third day and still going strong. Is there a
way to predict approximately how much time it will take computing a
regression? I mean - more than two days computational time deserves
some
kind of warning :)
Anyway - here's my printout, hope you can give me some advice.
Cheers,
Steinar Fossedal
. stset time, failure(default) id(kontrakt)
id: kontrakt
failure event: default != 0 & default < .
obs. time interval: (time[_n-1], time]
exit on or before: failure
------------------------------------------------------------------------
----
638684 total obs.
52310 obs. begin on or after (first) failure
------------------------------------------------------------------------
----
586374 obs. remaining, representing
31746 subjects
13582 failures in single failure-per-subject data
570537.2 total analysis time at risk, at risk from t = 0
earliest observed entry t = 0
last observed exit t = 68
. constraint 1 [kappa]_cons=0
. streg ,d(gamma) nolog time constraints(1)
failure _d: default
analysis time _t: time
id: kontrakt
Gamma regression -- accelerated failure-time form
No. of subjects = 31746 Number of obs =
586374
No. of failures = 13582
Time at risk = 570537.2198
Wald chi2(0) =
.
Log likelihood = -27747.888 Prob > chi2 =
.
------------------------------------------------------------------------
----
_t | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------
----
_cons | 3.534992 .0086277 409.73 0.000 3.518082
3.551902
-------------+----------------------------------------------------------
----
/ln_sig | -.3399564 .0129331 -26.29 0.000 -.3653049
-.3146079
/kappa | .8868395 .0246913 35.92 0.000 .8384454
.9352337
-------------+----------------------------------------------------------
----
sigma | .7118014 .0092058 .693985
.7300751
------------------------------------------------------------------------
----
*
* 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/
*
* 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/
