Timeline for Problem in finding a non-parametric confidence interval for median and mean using Frank Harrell approach when we have more than two categories
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 6, 2021 at 2:41 | answer | added | Yonghao | timeline score: 1 | |
| Jul 29, 2021 at 14:01 | comment | added | Rani | aha, so no one does the difference CI f? I used your suggestion in the textbook to do the CI for difference two mean on two group:>> diffs = numeric(2000) for(i in 1 : 20000){ diffs[i] = mean(sample(xsub1 , replace = TRUE )) - mean(sample(xsub2 , replace = TRUE )) } does this corerct? | |
| Jul 29, 2021 at 13:53 | comment | added | Frank Harrell |
The next release implements the delta method for getting better confidence intervals for means an quantiles. But that doesn't help with differences in means or quantiles. For now you'd need to put everything in a bootstrap loop to get bootstrap CLs, or use the Bayesian rms package rmsb.
|
|
| Jul 29, 2021 at 13:13 | comment | added | Rani | I use version 6.2-0, I want to learn this approach to do the CI for difference mean and difference median, not only for OR. I saw that I think in your book bbr. Could you please post more details with R for more clarity. | |
| Jul 29, 2021 at 11:49 | comment | added | Frank Harrell |
wilcox.test is using different (better) approach for CLs of differences: the Hodges-Lehmann estimator. This if for continuous Y (minimal ties) and is completely consistent with the WIlcoxon test. You'll have to run it in pairs since Kruskal-Wallis function doesn't do this. Which version of the rms package are you using? Also note that the rmsb package blrm function along with contrast and Quantile can provide exact (to within simulation error) Bayesian uncertainty intervals for a series of difference in means or quantiles using the proportional odds model.
|
|
| Jul 24, 2021 at 19:05 | review | First posts | |||
| Jul 25, 2021 at 3:16 | |||||
| Jul 24, 2021 at 19:00 | history | asked | Rani | CC BY-SA 4.0 |