Dear list, I made a logistic regression model (MyModel) using lrm and penalization by pentrace for data of 104 patients, which consists of 5 explanatory variables and one binary outcome (poor/good). Then, I found bootcov and robcov function in rms package for calculation of confidence range of coefficients and odds ratio by bootstrap covariance matrix and Huber-White sandwich method, respectively.
> MyModel.boot <- bootcov(MyModel, B=1000, coef.reps=T) > MyModel.robcov <- robcov(MyModel) > anova(MyModel) Wald Statistics Response: outcome Factor Chi-Square d.f. P stenosis 0.20 1 0.6547 x1 10.69 1 0.0011 x2 2.33 1 0.1270 procedure 3.27 1 0.0708 ClinicalScore 2.55 1 0.1102 TOTAL 18.71 5 0.0022 > anova(MyModel.boot) Wald Statistics Response: outcome Factor Chi-Square d.f. P stenosis 0.16 1 0.6921 x1 17.90 1 <.0001 x2 3.36 1 0.0669 procedure 4.62 1 0.0316 ClinicalScore 1.82 1 0.1774 TOTAL 31.82 5 <.0001 > anova(MyModel.robcov) Wald Statistics Response: outcome Factor Chi-Square d.f. P stenosis 0.17 1 0.6758 x1 20.52 1 <.0001 x2 3.83 1 0.0505 procedure 5.09 1 0.0241 ClinicalScore 1.84 1 0.1744 TOTAL 34.80 5 <.0001 The confidence intervals are narrower in bootcov model, and further narrower in robcov model than in original model, as demonstrated in the followings. I am wondering which confidence interval should be used. I would appreciate anybody's help in advance. -- KH > summary(MyModel, stenosis=c(70, 80), x1=c(1.5, 2.0), x2=c(1.5, 2.0)) Effects Response : outcome Factor Low High Diff. Effect S.E. Lower 0.95 Upper 0.95 stenosis 70.0 80 10.0 -0.11 0.24 -0.59 0.37 Odds Ratio 70.0 80 10.0 0.90 NA 0.56 1.45 x1 1.5 2 0.5 1.21 0.37 0.49 1.94 Odds Ratio 1.5 2 0.5 3.36 NA 1.63 6.95 x2 1.5 2 0.5 -0.29 0.19 -0.65 0.08 Odds Ratio 1.5 2 0.5 0.75 NA 0.52 1.08 ClinicalScore 3.0 5 2.0 0.61 0.38 -0.14 1.36 Odds Ratio 3.0 5 2.0 1.84 NA 0.87 3.89 procedure - CA:CE 2.0 1 NA 0.83 0.46 -0.07 1.72 Odds Ratio 2.0 1 NA 2.28 NA 0.93 5.59 > summary(MyModel.boot, stenosis=c(70, 80), x1=c(1.5, 2.0), x2=c(1.5, 2.0)) Effects Response : outcome Factor Low High Diff. Effect S.E. Lower 0.95 Upper 0.95 stenosis 70.0 80 10.0 -0.11 0.28 -0.65 0.43 Odds Ratio 70.0 80 10.0 0.90 NA 0.52 1.54 x1 1.5 2 0.5 1.21 0.29 0.65 1.77 Odds Ratio 1.5 2 0.5 3.36 NA 1.92 5.89 x2 1.5 2 0.5 -0.29 0.16 -0.59 0.02 Odds Ratio 1.5 2 0.5 0.75 NA 0.55 1.02 ClinicalScore 3.0 5 2.0 0.61 0.45 -0.28 1.50 Odds Ratio 3.0 5 2.0 1.84 NA 0.76 4.47 procedure - CAS:CEA 2.0 1 NA 0.83 0.38 0.07 1.58 Odds Ratio 2.0 1 NA 2.28 NA 1.08 4.85 > summary(MyModel.robcov, stenosis=c(70, 80), T1=c(1.5, 2.0), T2=c(1.5, 2.0)) Effects Response : outcome Factor Low High Diff. Effect S.E. Lower 0.95 Upper 0.95 stenosis 70.0 80 10.0 -0.11 0.26 -0.62 0.40 Odds Ratio 70.0 80 10.0 0.90 NA 0.54 1.50 x1 1.5 2 0.5 1.21 0.27 0.69 1.74 Odds Ratio 1.5 2 0.5 3.36 NA 1.99 5.68 x2 1.5 2 0.5 -0.29 0.15 -0.57 0.00 Odds Ratio 1.5 2 0.5 0.75 NA 0.56 1.00 ClinicalScore 3.0 5 2.0 0.61 0.45 -0.27 1.49 Odds Ratio 3.0 5 2.0 1.84 NA 0.76 4.44 procedure - CAS:CEA 2.0 1 NA 0.83 0.37 0.11 1.54 Odds Ratio 2.0 1 NA 2.28 NA 1.11 4.68 ______________________________________________ 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.