Fig. 3.

A, B Heat maps displaying the absolute bias (Monte Carlo standard error) of each meta-regression model, for each scenario and age overlap for the linear age-varying genetic effect (${\beta }_{SNP\times Age}$)(N = 1000). Squares represent scenarios where ${\beta }_{SNP\times Age}=0.02$ (scenario 2&3) and ${\beta }_{SNP\times Age}=0.04$ (scenario 5). In other scenarios${\beta }_{SNP\times Age}=0$. C Heat maps displaying the absolute bias (Monte Carlo standard error) of meta-regression including a non-linear term of age, for each scenario and age overlap for the non-linear age-varying genetic effect (${\beta }_{{SNP\times age}^2}$) (N = 1000). In scenario 5${\beta }_{{SNP\times age}^2}=0.001$. In all other scenarios${\beta }_{{SNP\times age}^2}=0$