Determining if the number of significant results between two categories is significantly different?
0
0
Entering edit mode
4.0 years ago
DNAngel ▴ 250

Hi all,

Quick concept question here...

If I have two groups ("Functional" with 100 genes, "Control" with 50 genes) and I ran a bunch of models on each gene individually and found that "Functional" out of 100 genes, 50 were significant for a test, and "Control" (although I expected none) I got 10 showing a significant result.

Now, I still want to know if the proportion of significant results is different between these two unequal groups as that can still be meaningful to me.

Do I compare every single actual p-value (100 p-values in Functional, 50 in Control) using a one-tailed t-test (because I originally only expected significant results in the Functional category, so now I am expecting it to at least have a greater number of significant tests than in Control). OR...should I convert the p-values to remove any variation so any p-value <= 0.05, I change to 0.05 and anything above this I change to 1. I am thinking to do this to avoid really small/large p-values from driving variation in this test as I essentially want to know if the proportion of significant results is sig diff between the two groups. Does this make sense?

statistics • 643 views
ADD COMMENT
0
Entering edit mode

This can be answered with a hypergeometric test.

ADD REPLY
0
Entering edit mode

Sorry for my ignorance but this would be different from a proportions test right? (Z-test)?

ADD REPLY

Login before adding your answer.

Traffic: 1913 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6