tangram
tangram copied to clipboard
multiple comparison correction support
Hi Shawn,
I find this package very handy to create tables for reports, especially it provides statistical tests and reports p-values back.
But my question is, are the reported p-values corrected for the multiple comparison problem? if it is, what method was used?
Thanks,
Yang
Assuming you question refers to the default rmisc transform the answer is no correction is applied.
The rmisc transform is our departments recommended exploratory summary of data. The p-values are simply to look for something 'worthy of further investigation'.
Also a transform has no knowledge of the broader number of methods being applied. There is support in the hmisc module for specifying ones own statistical model to generate p-values. You're free to do whatever method you desire.
I spoke with Frank Harrell and these p-values have caused more confusion and aren't worthy of reporting in general. He's now recommending that they be removed entirely.