Dear Martin,
I tried to load the excerpt of the data set provided in your previous message. Whereas Cox proportional hazard model seems to hold (please, see -estat phtest-), the Gamma generalized regression for finding which parametric survival model fits best with your data, doesn't converge (please, see the error message).
stset year, id(id) failure(health_state==1) exit(time .) scale(1)
id: id
failure event: health_state == 1
obs. time interval: (year[_n-1], year]
exit on or before: time .
------------------------------------------------------------------------------
21 total obs.
0 exclusions
------------------------------------------------------------------------------
21 obs. remaining, representing
3 subjects
6 failures in multiple failure-per-subject data
21 total analysis time at risk, at risk from t = 0
earliest observed entry t = 0
last observed exit t = 7
stcox female weight, schoenfeld(sch*)
failure _d: health_state == 1
analysis time _t: year
exit on or before: time .
id: id
Iteration 0: log likelihood = -5.7807435
Iteration 1: log likelihood = -4.7542542
Iteration 2: log likelihood = -4.7462302
Iteration 3: log likelihood = -4.746223
Refining estimates:
Iteration 0: log likelihood = -4.746223
Cox regression -- no ties
No. of subjects = 3 Number of obs = 18
No. of failures = 6
Time at risk = 18
LR chi2(2) = 2.07
Log likelihood = -4.746223 Prob > chi2 = 0.3554
------------------------------------------------------------------------------
_t | Haz. Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
female | .2556673 .3192123 -1.09 0.275 .0221264 2.954196
weight | 1.079909 .0695928 1.19 0.233 .9517721 1.225297
------------------------------------------------------------------------------
. estat phtest
Test of proportional hazards assumption
Time: Time
----------------------------------------------------------------
| chi2 df Prob>chi2
------------+---------------------------------------------------
global test | 2.77 2 0.2505
----------------------------------------------------------------
streg female weight, dist(gamma)
failure _d: health_state == 1
analysis time _t: year
exit on or before: time .
id: id
<output omitted>
discontinuous region encountered
cannot compute an improvement
r(430);
I cannot rule out the occurrence of some mistakes in my tackling the issues you outlined.
I do hope that it (at least a bit) help.
Kind Regards,
Carlo
-----Messaggio originale-----
Da: [email protected] [mailto:[email protected]] Per conto di Martin Wang
Inviato: lunedì 3 novembre 2008 17.03
A: [email protected]; [email protected]
Oggetto: Re: st: R: Someone help me on the multiple failure duration model?
Dear Carlo,
Many thanks for your suggestion. I've just requested the book from other library. But my question is a little bit urgent, if anyone happens to know the answer could you please kindly help.
Thanks!
Martin
======= 2008-11-03 23:40:52 Original Message：=======
>Dear Martin,
>all these topics are covered in Cleves MA, Gould WW, Gutierrez RG. An
>introduction to survival analysis using Stata. Revised edition. College
>Station: Stata Press, 2004.
>
>In my opinion, this is a very "good value for money" textbook.
>
>
>Kind Regards,
>
>Carlo
>-----Messaggio originale-----
>Da: [email protected]
>[mailto:[email protected]] Per conto di Martin Wang
>Inviato: luned?3 novembre 2008 16.22
>A: [email protected]
>Oggetto: st: Someone help me on the multiple failure duration model?
>
>/********** This is a html email************************/
>
>Dear all,
>
>Can someone help me on the multiple failure duration model? For example,
>let's say I have the following data:
>
>
>year | 1 | 2 | 3 |
>4 | 5 | 6 | 7
>patient 1(Female)|weight:50,ill(0)| weight:52,ill(0) |
>weight:53,fine(1)|weight:48,fine(1)|weight:50,ill(0) |weight:50,ill(0)
>|weight:50,ill(0)
>patient 2(male) |weight:51,ill(0)| weight:48,ill(0) | weight:60,ill(0) |
>weight:67,ill(0)|weight:78,fine(1) |weight:60,fine(1)| (missing)
>patient 3(male) |weight:52,ill(0)| weight:50,fine(1)| (missing) |
>(missing) | weight:50,fine(1)| weight:40,ill(0)| weight:60,ill(0)
>
>
>You see for each patient there are multiple observations, and also there may
>be multiple failures(in this case - getting fine). And also some subject has
>gap in the data, for example patient 3 has gap in year 3 and year 4. I
>wonder whether I can still use streg or stcox for this setting and use ALL
>the observations. For example,
>
>*****************************************************
>stset t, id(patient) failure(fine) exit(time .) noshow
>stcox weight female, r
>or streg weight female, r
>*****************************************************
>Can anyone kindly help me out? Any suggestions?
>
>Many thanks!!
>
>Martin
>
>*
>* For searches and help try:
>* http://www.stata.com/help.cgi?search
>* http://www.stata.com/support/statalist/faq
>* http://www.ats.ucla.edu/stat/stata/
>
>
>
>*
>* For searches and help try:
>* http://www.stata.com/help.cgi?search
>* http://www.stata.com/support/statalist/faq
>* http://www.ats.ucla.edu/stat/stata/
= = = = = = = = = = = = = = = = = = = =
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
