Is there any way to make p-value high if correlation has negative value?

Question

[Simon Ahn]

Hi. I'm trying to figure out the p-value of correlation between CNV and mRNA expression.

If the chromosome is amplified, expression of mRNA involving that chromosome is likely to be increased.

On the other hand, if the chromosome suffered deletion, expression of mRNA involving that chromosome is likely to be decreased.

In other words, positive correlation is expected result.

I have CNV data and mRNA expression level data and calculated correlation for each genes.

Obviously, if correlation value is high, p-value is low in possitive correlation.

But, negative correlations also have row p-value and I don't want that.

So, in summary, I want to give high p-value if one gene has negative correlation. (It shouldn't be p-value)

What method should I use?

Answer 1

[dsull]

I'm assuming you're either performing a permutation test or a t test to get your p-values.

Use a one-sided test rather than a two-sided test. The one-sided p-value is obtained by simply dividing the two-sided p-value (p) by 2 (aka p/2) if your correlation is positive and, if your correlation is negative, the one-sided p-value is 1 - p/2.

[Simon Ahn]

This is exactly what I wanted! Thanks a lot!