My apologies for asking slightly about SPSS in addition to R... Could not find an exact answer in the archives on whether R and SPSS may give different p-vals when output for coeffs and conf-intervals are the same. Amyway, a colleague and I are doing a very simple coxreg analyses and get the same results for the coefficient and confidence interval, exp(coef) exp(-coef) lower .95 upper .95 age_at_entry 1.02 0.98 1.01 1.03 but in R we get p = 0.00011, and SPSS gives p < 0.0001 Should we worry about this difference in p-value or do R and SPSS sometime differ? All the best, Kare
K?re Edvardsen wrote:> My apologies for asking slightly about SPSS in addition to R... > > Could not find an exact answer in the archives on whether R and SPSS may > give different p-vals when output for coeffs and conf-intervals are the > same. > Amyway, a colleague and I are doing a very simple coxreg analyses and > get the same results for the coefficient and confidence interval, > > exp(coef) exp(-coef) lower .95 upper .95 > age_at_entry 1.02 0.98 1.01 1.03 > > > but in R we get p = 0.00011, and SPSS gives p < 0.0001This might happen due to numerical differences in the algorithms for computing on the distributions. Both p values are not that different ... For I would not worry if the difference is that small and practically identical. Uwe Ligges> Should we worry about this difference in p-value or do R and SPSS > sometime differ? > > All the best, > Kare > > ______________________________________________ > R-help at 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.
That is a larger difference in p-values than I would expect due to numerical differences and stopping criteria. My guess is that you are running across the different approximations for tied failure times. If so, you will get better agreement with SPSS by using method="breslow" in coxph(). -thomas On Thu, 11 Sep 2008, K?re Edvardsen wrote:> My apologies for asking slightly about SPSS in addition to R... > > Could not find an exact answer in the archives on whether R and SPSS may > give different p-vals when output for coeffs and conf-intervals are the > same. > Amyway, a colleague and I are doing a very simple coxreg analyses and > get the same results for the coefficient and confidence interval, > > exp(coef) exp(-coef) lower .95 upper .95 > age_at_entry 1.02 0.98 1.01 1.03 > > > but in R we get p = 0.00011, and SPSS gives p < 0.0001 > > Should we worry about this difference in p-value or do R and SPSS > sometime differ? > > All the best, > Kare > > ______________________________________________ > R-help at 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. >Thomas Lumley Assoc. Professor, Biostatistics tlumley at u.washington.edu University of Washington, Seattle