Re: [R] Cox Proportional Hazards model with a time-varying covariate

[Paul Miller]

Hi Terry and Kevin,
 
Thanks for your replies to my post. And thanks particularly to Terry for 
pointing out my confusion about the "exact partial likelihood" vs. the "exact 
marginal likelihood." I'm learning and so am likely to be confused about such 
things from time to time. 
 
I had thought that a good way to see if I was replicating the results would be 
to use the Efron approximation in both R and SAS. I tried this just now and 
obtained a similar pattern of differences in the results. Please bear with me. 
I hope I'm not doing something silly.
 
Below are my results in each program. I got the results for R simply by 
running the code I sent earlier. I got the results in SAS by changing TIES 
= EXACT to TIES = EFRON.
 
Should I be getting results as different as these? Or does this suggest that 
there is some problem in how I've set up my data in R? I've looked over my 
data In R and everything looks fine, but maybe I've missed something.
 
Thanks,
 
Paul
 
> summary(model)
Call:
coxph(formula = Surv(START, STOP, STATUS) ~ TRT + DENS + INFCTN, 
    data = vitclear)
  n= 167 
          coef exp(coef) se(coef)      z Pr(>|z|)
TRT     0.2913    1.3382   0.1782  1.635    0.102
DENS   -0.1762    0.8384   0.1739 -1.013    0.311
INFCTN  0.1192    1.1266   0.1912  0.624    0.533
       exp(coef) exp(-coef) lower .95 upper .95
TRT       1.3382     0.7473    0.9436     1.898
DENS      0.8384     1.1927    0.5963     1.179
INFCTN    1.1266     0.8876    0.7746     1.639
Rsquare= 0.024   (max possible= 0.999 )
Likelihood ratio test= 4.12  on 3 df,   p=0.2486
Wald test            = 4.12  on 3 df,   p=0.2485
Score (logrank) test = 4.15  on 3 df,   p=0.2460
 
 
Cox Proportional Hazards Model 09:01 Friday, April 1, 2011 1
Example 22.2: Hyalurise in Vitreous Hemorrhage
The PHREG Procedure
Model Information
Data Set WORK.VITCLEAR
Dependent Variable RSPTIM
Censoring Variable CENS
Censoring Value(s) 0
Ties Handling EFRON
　
Number of Observations Read 143
Number of Observations Used 143
　
Summary of the Number of Event and Censored Values
Percent
Stratum CENTER Total Event Censored Censored
1 A 60 45 15 25.00
2 B 37 30 7 18.92
3 C 46 39 7 15.22
-------------------------------------------------------------------
Total 143 114 29 20.28
　
Convergence Status
Convergence criterion (GCONV=1E-8) satisfied.
　
　
　
Cox Proportional Hazards Model 09:01 Friday, April 1, 2011 2
Example 22.2: Hyalurise in Vitreous Hemorrhage
The PHREG Procedure
Model Fit Statistics
Without With
Criterion Covariates Covariates
-2 LOG L 711.142 706.171
AIC 711.142 712.171
SBC 711.142 720.379
　
Testing Global Null Hypothesis: BETA=0
Test Chi-Square DF Pr > ChiSq
Likelihood Ratio 4.9712 3 0.1739
Score 4.9604 3 0.1747
Wald 4.9151 3 0.1781
　
Analysis of Maximum Likelihood Estimates
Parameter Standard Hazard
Parameter DF Estimate Error Chi-Square Pr > ChiSq Ratio
TRT 1 0.34437 0.19562 3.0988 0.0783 1.411
DENS 1 -0.23832 0.19585 1.4807 0.2237 0.788
INFCTN 1 0.12819 0.27780 0.2129 0.6445 1.137
　
 
        [[alternative HTML version deleted]]


______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.