Figure 2.
Distributions of regression estimates resulting from leave-one-out permutation analysis. Solid line = estimate from main analysis (n=96 variants); dashed line = estimate with TCF7L2(rs7903146) removed; dotted line = estimate with FTO(rs1558902) removed.
A - Causal estimates (β coefficients) of BMI on type 2 diabetes estimated by an inverse-variance weighted method
B - Causal estimates (β coefficients) of BMI on type 2 diabetes estimated by MR-Egger
C - Causal estimates (β coefficients) of BMI on type 2 diabetes estimated by a weighted median method
D – Estimates of the intercept by MR-Egger