Figure 4. Arbitrary combining of length and angle tests.
We conducted a simulation analysis and used W(W1, p, q) with p=q=2, and let w1 = 1 (length only), w1 = 0.75, w1 =0.5 (typical joint test), w1 =0.25 and w10 (angle only test). Figure 4 illustrates the power of these five tests across different values of λ2. Power curves are as predicted. In particular, the length only (w1 =1) and angle only (w1 =0) tests show the least robustness, while the (w1 =0.5) test is quite robust. As expected, the (w1 =0.75) weighted test outperforms the (w1 =0.5) test when both variants are risk-inducing, while providing more power than the length only test when there is a mix of risk-inducing and protective variants. The reverse is true for the (w1 =0.25) test.