| Method | Description | Weight | Test Statistic |
---|---|---|---|---|
General method | Score Test \(S_s\) | Maximum of the weighted sum of Z scores | \(\varvec{W }=\varvec{ Z}^{\varvec{'}} \varvec{R}^{\varvec{-1}}\) | \(S_{s} = \varvec{Z}^{\varvec{'}}\varvec{R}^{\varvec{-1}}\varvec{Z}\) |
Special cases of the score test \(S_s\) | Special case 1: Sum of Squared Score Statistic (SSU) | Weighted sum of Z scores, weights are Z scores | \(\varvec{W =R^{-1}Z, R=I}\) | \(S_{Q} = \varvec{Z^{'}Z}\) |
Special case 2: SNP-set (Sequence) Kernel Association Test (SKAT) | Weighted sum of Z scores, weights are weighted Z scores | \(\varvec{W =R^{-1}Z}, \varvec{R}=diag(a_{1}, \ldots , a_{m}), a_{m} \sim beta(1,25)\) | \(S_{SKAT} = \varvec{Z^{'}R^{-1}Z}\) | |
Special case 3: PathSPU(2) | Weighted sum of Z scores, weights are eQTL weighted Z scores | \(\varvec{W =R^{-1}Z}, \varvec{R}=diag(a_{1}, \ldots , a_{m}), a_{m}\) are gene derived weights | \(S_{pathSPU(2)} = \varvec{Z^{'}R^{-1}Z}\) | |
Special case 4: Sum of Powered Score (SPU): Data-adaptive weighted combination test. | Weighted sum of Z scores, weights are function of Z scores | \(\varvec{W} = \varvec{Z}^{\gamma - 1}\) | \(SPU(\gamma ) = \sum _{m=1}^{M}\varvec{Z}_m^{\gamma }, \gamma = 1,2,\ldots ,8,\infty\) | |
Special case 5: Burden test | Weighted sum of Z scores, weights are all 1s | \(\varvec{W} = (1, \ldots , 1)^{'}\) | \(L_B = L(1, \ldots , 1) = \sum _{m=1}^{M}{\textbf{Z}}_{\textbf{m}}\) | |
Special case 6: Weighted Sum Statistic | Weighted sum of Z scores, weights are related to MAFs | \(\varvec{W} = (\frac{1}{\sqrt{p_1(1-p_1)}}, \ldots , \frac{1}{\sqrt{p_m(1-p_m)}})^{'}\), where \(p_m\) is the MAF | \(L_W = L(\frac{1}{\sqrt{p_1(1-p_1)}}, \ldots , \frac{1}{\sqrt{p_m(1-p_m)}}) = \sum _{m=1}^{M}\frac{1}{\sqrt{p_m(1-p_m)}}\cdot {\textbf{Z}}_{\textbf{m}}\) |