Skip to content
Aaron Stern edited this page May 6, 2020 · 10 revisions

Frequently asked questions

Q: How do I choose the P-value threshhold for SNPs to include in my analysis?

A: Because the current version of PALM does not explicitly account for uncertainty in effect size estimates, we recommend using a roughly genome-wide significant (GWS) P-value threshhold. Using a threshhold also has the advantage of reducing the relative bias due to uncorrected GWAS stratification in your analyses.


Q: How do I interpret estimates of the selection gradient (Sel and/or Z)?

A: Assuming no confounders (i.e., all pleiotropically related traits under direct selection are included in the PALM analysis), the estimate of the selection gradient is a measure of direct selection.

However, realistically you cannot include all genetically correlated traits. Hence, in realistic settings it is actually a measure of indirect selection, but controls for pleiotropy to the other traits in the analysis.

The selection gradient measures the increase in fitness associated with a unit increase in the trait value.


Q: How do I interpret estimates of correlated response (R)?

A: Assuming no confounders (same definition as above), R can be viewed as a measure of how much indirect selection the trait experiences due to direct selection on the other traits in the analysis.

For example, if Trait I is under positive selection, Trait II is neutral, and rg >> 0%, then MARGINALLY Z_I >> 0, Z_II >> 0, JOINTLY Z_I >> 0, Z_II ~ 0, and R_I ~ 0, R_II >> 0.

Clone this wiki locally