Figure 1.

PolyMR is able to recover the shape of the causal function

The true causal function is shown in green (solid line). The observed association model is shown in orange (short dashed) while that obtained using PolyMR is shown in purple (long dashed). The hulls around the model curves show the 95% coverage hull across 1,000 simulations. The hulls around the observational effects represent the 95% confidence interval estimated from the conditional distribution of the outcome. The y axis shows the expected association with/effect of the exposure on the outcome, relative to the outcome level at the mean population exposure. (A) A setting representing high polygenicity (10,000 causal SNPs accounting for a heritability of 0.3; see Table 1 for details). (B) Results for a sigmoid causal effect ($f_{\alpha }\left(\mathsf{X}\right)=0.1\cdot \frac{1}{1+{\mathsf{e}}^{-2\cdot \mathsf{X}}}$ scenario.