|
I made 100 comparisons by applying chi-square test (tajima relative rate test) and got 100 p values. Do I need P-value adjustment? If yes then which one correction will be suitable? I have explored internet and also on this forum but didnt get answer of this particular question. |
|||
|
|
|
I would have thought you definitely need an adjustment. It's difficult to give advice without knowing more about the problem. If you are using the p-value as a guide to whether there is a real effect or not, you would expect 5 out of 100 p-values to be less than 0.05 if there is actually no effect/relationship at all. If you have a lot more "significant" p-values than 5, that may be an indicator that there is something interesting in the data, but it also depends on how highly correlated all your 100 outcome measures are. |
|||
|
|
|
There are many acceptable ways to do p-value adjustment. But some like Bonferroni are more conservative than others. Because you are doing so many pairwise tests you don't want conservative tests. In fact you may want to think about whether you want the criterion be control of FWER type I error or control of FDR. When large numbers of hypotheses are being tested it is common these days to look at false detection rate (FDR). Resampling p-value adjustment as described in the book by Westfall and Young is an approach that I would recommend. |
|||
|
|
