Figure 3.

Scatter plot of respective effect size of each SNP for the use of sun/UV protection (A) and ease of skin tanning (B) on PXG. (1) Inverse variance weighted (A1 & B1), a primary regression with Wald ratios estimated for each SNP, meta-analyzed under a fixed-effects model; (2) maximum likelihood (A2 & B2), which uses a maximum likelihood method where the genetic effects on the exposure and outcome are modeled directly as a bivariate normal distribution; (3) weighted median (A3 & B3), which provides robust point estimates even when up to 50% of the IVs are invalid instruments; (4) MR Egger (A4 & B4), which accounts for directional pleiotropy.