[S] Cox PH warning

Pintilie, Melania (Melania_Pintilie@pmh.toronto.on.ca)
Wed, 2 Sep 1998 09:59:06 -0400


Hi,
I want to use stepwise selection technique for Cox proportional hazard
model. I ran the following lines of program:
y_coxph(Surv(dftime,dfcens)~stage+rt+surgcat, data=subset
y2_stepAIC(y,scope=list(lower=~1,
upper=~stage+rt+surgcat+sizecat+adher+rupt+grcat+age),
trace=1,direction="both")

I got the following results:
Start: AIC= 58.12
Surv(dftime, dfcens) ~ stage + rt + surgcat

df AIC
+grcat 4 54.75588
+rupt 4 57.25402
<none> 3 58.12332
+adher 4 58.61405
-surgcat 2 58.62017
+sizecat 4 59.83129
+age 4 60.11220
-rt 2 61.38828
-stage 2 63.84060

Step: AIC= 54.76
Surv(dftime, dfcens) ~ stage + rt + surgcat + grcat
df AIC
-surgcat 3 52.93321
<none> 4 54.75588
+rupt 5 55.08867
+adher 5 55.96370
-rt 3 56.56491
+age 5 56.72469
+sizecat 5 56.75348
-grcat 3 58.12332
-stage 3 63.71189

Step: AIC= 52.93
Surv(dftime, dfcens) ~ stage + rt + grcat
Surv(dftime, dfcens) ~ stage + rt + grcat

df AIC
<none> 3 52.93321
+rupt 4 53.08882
+adher 4 54.28501
-rt 2 54.60309
+surgcat 4 54.75588
+sizecat 4 54.86546
+age 4 54.93094
-grcat 2 58.62017
-stage 2 61.74631
df AIC
-surgcat 3 52.93321
<none> 4 54.75588
+rupt 5 55.08867
+adher 5 55.96370
-rt 3 56.56491
+age 5 56.72469
+sizecat 5 56.75348
-grcat 3 58.12332
-stage 3 63.71189

Step: AIC= 52.93
Surv(dftime, dfcens) ~ stage + rt + grcat
Surv(dftime, dfcens) ~ stage + rt + grcat

df AIC
<none> 3 52.93321
+rupt 4 53.08882
+adher 4 54.28501
-rt 2 54.60309
+surgcat 4 54.75588
+sizecat 4 54.86546
+age 4 54.93094
-grcat 2 58.62017
-stage 2 61.74631
df AIC
-surgcat 3 52.93321
<none> 4 54.75588
+rupt 5 55.08867
+adher 5 55.96370
-rt 3 56.56491
+age 5 56.72469
+sizecat 5 56.75348
-grcat 3 58.12332
-stage 3 63.71189

Step: AIC= 52.93
Surv(dftime, dfcens) ~ stage + rt + grcat
Surv(dftime, dfcens) ~ stage + rt + grcat

df AIC
<none> 3 52.93321
+rupt 4 53.08882
+adher 4 54.28501
-rt 2 54.60309
+surgcat 4 54.75588
+sizecat 4 54.86546
+age 4 54.93094
-grcat 2 58.62017
-stage 2 61.74631

and 15 warnings like this:

Warning messages:
1: Loglik converged before variable 1 ,4; beta may be infinite. in:
fitter(X,
Y, strats, offset, init = init, iter.max = iter.max, eps = eps,
....
2: Loglik converged before variable 1 ,5; beta may be infinite. in:
fitter(X,
Y, strats, offset, init = init, iter.max = iter.max, eps = eps,
....
3: Loglik converged before variable 1 ,4; beta may be infinite. in:
fitter(X,
Y, strats, offset, init = init, iter.max = iter.max, eps = eps,
....

QUESTION:
What do these warnings mean? Can I ignore them?

Your help is greatly appreciated.

Melania Pintilie
PMH Biostatistics


-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news