1
$\begingroup$

Suppose I have a test with multiple hypotheses.

For example, if I have X as a gene set (a set of symbols), and A, B, C pathways (sets of symbols), I am testing if A is enriched in X, B in X, C in X and so on.

Then I am correcting the resulting pvalues for multiple hypothesis testing.

Now I have several corrected pvalues, and I want to combine them to a single pvalue.

Is this possible?

I will add more info if needed.

$\endgroup$

1 Answer 1

1
$\begingroup$

A p-value is for a test of a particular hypothesis. There is (or should be) one test per hypothesis.

If you have a multiple regression, then you will get both an overall p value and one for each independent variable. However, each of these tests a different hypothesis.

Can you combine the p-values? Sure. You can take the mean, the median, whatever you like. But what will the result mean? If you can come up with a reasonable and substantively meaningful interpretation for this value, then go ahead and do it. But it's not a typical thing to do.

$\endgroup$
2
  • $\begingroup$ Hi Peter. Thanks for your answer. Sorry for the dumb question: what do you mean when you said multiple regression? I searched but having not a strong background on inference, I am having some problems understanding it. How does the multiple regression relate to the multiple pvalues I have in input? Thanks $\endgroup$
    – gc5
    May 29, 2014 at 9:29
  • $\begingroup$ Multiple regression makes one equation, so it has one p value for the entire equation. But it's a different test because all the IVs are controlled for. $\endgroup$
    – Peter Flom
    May 29, 2014 at 10:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.