| imsl_survival_glm [message #63517] |
Fri, 07 November 2008 05:15 |
Tom Over
Messages: 5
Registered: April 1997
|
Junior Member |
|
|
Has anyone used imsl_survival_glm much? I can run the example in the
help, but for my data, it doesn't want to output the coefficients
(keyword coef_stat), without which it is pretty useless.
The specific situation is an input array surv_x:
IDL> print, surv_x, format = "(16f10.2)"
40.29 246.14 230.09 1033.99 1.30 1326.68 520.64 2548.12 6359.52 103.44
8201.34 8754.44 145.24 157.74 164.67 9554.00
5.30 7.00 8.00 145.00 0.01 153.00 70.00 370.00 670.00 8.00 1000.00
715.00 12.00 13.00 12.00 1000.00
Then run imsl_survival_glm as follows:
IDL> ncoef = imsl_survival_glm(0,1,2,surv_x)
(no class variables, 1 continuous variable, lognormal model)
which appears to work fine (ncoef = 3)
But if I really want some output:
IDL> ncoef = imsl_survival_glm(0,1,2,surv_x,case_anal=ca,coef_stat=cs)
then I get the error:
"Array has a corrupted descriptor: COEF_STAT_SPC"
and IDL crashes (this is in IDL 6.4; the same IMSL routine from IDL7
behaves the same.)
Yes, this reduces to simple linear regression with x and y log-
transformed (according to my understanding anyway), but in reality the
5th y-value (0.01) is left-truncated, so I really intend to make use
of this routine for that, which is how I originally tried to run it,
but got the same error.
Any thoughts?
Thanks a lot,
Tom
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
