The "n = 1404" vs. "Number of obs = 468" looks like the giveaway.
You are passing the subset selection logic as the 3rd positional argument, but according to the documentation, that is the weights argument. So, clogit(..., data = dframe, subset = sample==1 & glb_ind=="Y") On Jun 7, 2013, at 18:51 , Richard Beckett wrote: > > > From: peter dalgaard <pda...@gmail.com> > To: Richard Beckett <rbecket...@yahoo.com> > Cc: "r-help@r-project.org" <r-help@r-project.org> > Sent: Friday, June 7, 2013 11:12 AM > Subject: Re: [R] Clogit R and Stata > > Here is the R output: > > Call: > coxph(formula = Surv(rep(1, 1404L), sftpcons) ~ sftptv2a3 + sftptv2a4 + > sftptv2a5 + sftptv2a2 + sftptv2a6 + logim + maccat + disp4cat + > strata(stratida), data = dframe, method = "exact") > > n= 1404, number of events= 351 > > coef exp(coef) se(coef) z Pr(>|z|) > sftptv2a3 1.4552 4.2852 0.2273 6.401 1.54e-10 *** > sftptv2a4 3.1118 22.4609 0.2265 13.739 < 2e-16 *** > sftptv2a5 1.0717 2.9204 0.2522 4.249 2.15e-05 *** > sftptv2a2 0.7185 2.0514 0.3300 2.177 0.0295 * > sftptv2a6 2.7341 15.3965 0.5050 5.414 6.17e-08 *** > logim 0.7579 2.1338 0.1347 5.625 1.85e-08 *** > maccat 3.0809 21.7771 0.4005 7.693 1.43e-14 *** > disp4cat 0.7061 2.0261 0.1524 4.634 3.59e-06 *** > --- > Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > > exp(coef) exp(-coef) lower .95 upper .95 > sftptv2a3 4.285 0.23336 2.745 6.691 > sftptv2a4 22.461 0.04452 14.409 35.013 > sftptv2a5 2.920 0.34241 1.781 4.788 > sftptv2a2 2.051 0.48747 1.074 3.917 > sftptv2a6 15.397 0.06495 5.722 41.429 > logim 2.134 0.46866 1.639 2.779 > maccat 21.777 0.04592 9.934 47.739 > disp4cat 2.026 0.49355 1.503 2.731 > > Rsquare= 0.239 (max possible= 0.623 ) > Likelihood ratio test= 383.2 on 8 df, p=0 > Wald test = 264.7 on 8 df, p=0 > Score (logrank) test = 396.2 on 8 df, p=0 > > > And the STATA output: > > Iteration 0: log likelihood = -95.537697 > Iteration 1: log likelihood = -91.465581 > Iteration 2: log likelihood = -91.402366 > Iteration 3: log likelihood = -91.402264 > Iteration 4: log likelihood = -91.402264 > > Conditional (fixed-effects) logistic regression Number of obs = 468 > LR chi2(8) = 141.59 > Prob > chi2 = 0.0000 > Log likelihood = -91.402264 Pseudo R2 = 0.4365 > > ------------------------------------------------------------------------------ > sftpcons | Coef. Std. Err. z P>|z| [95% Conf. Interval] > -------------+---------------------------------------------------------------- > sftptv2a3 | 2.042827 .4741327 4.31 0.000 1.113544 2.97211 > sftptv2a4 | 4.10828 .5593723 7.34 0.000 3.01193 5.204629 > sftptv2a5 | 1.766492 .5585173 3.16 0.002 .6718177 2.861165 > sftptv2a2 | 1.366568 .6540307 2.09 0.037 .084691 2.648444 > sftptv2a6 | 2.307152 .8225835 2.80 0.005 .6949178 3.919386 > logim | 1.404135 .3480976 4.03 0.000 .7218764 2.086394 > maccat | 2.8423 .7008588 4.06 0.000 1.468642 4.215958 > disp4cat | .6347805 .2872258 2.21 0.027 .0718283 1.197733 > ------------------------------------------------------------------------------ > > Also tried changing method=approximate with no noticeable change > > On Jun 7, 2013, at 15:34 , Richard Beckett wrote: > > > Sorry to once again write a message but I'm once again stumped and am > > having no luck finding a solution anywhere else. > > > > > > This question requires some finesse in both R and STATA so hopefully I will > > be able to get an answer here. I am much more adept in R and am trying to > > replicate the results of a STATA file in R. Hopefully this is a proper > > forum for such questions. > > > > > > This is the code for the clogit in STATA > > clogit sftpcons sftptv2a3 sftptv2a4 sftptv2a5 sftptv2a2 sftptv2a6 logim > > maccat disp4cat if sample==1 & glb_ind=="Y", group(stratida) > > and I tried to replicate it using > > clogit1<-clogit(sftpcons~sftptv2a3+sftptv2a4+sftptv2a5+sftptv2a2+sftptv2a6+logim+maccat+disp4cat+strata(stratida), > > dframe, sample==1 | glb_ind=="Y") > > but got different results > > What did I do wrong here? I interpreted the STATA clogit as run this logit > > as long as the sample is 1 and glb_ind="Y" What should I be doing instead? > > > An "&" rather than "|" in the R version might help. Other than that, we're a > bit short on clues unless you provide some output. > > -- > Peter Dalgaard, Professor, > Center for Statistics, Copenhagen Business School > Solbjerg Plads 3, 2000 Frederiksberg, Denmark > Phone: (+45)38153501 > Email: pd....@cbs.dk Priv: pda...@gmail.com > > > > > > > > > > -- Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Email: pd....@cbs.dk Priv: pda...@gmail.com ______________________________________________ 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.
