Figure 4.

Power when a Subset of Phenotypes Are Associated with SNP

Power results for control-only analysis $\left(n_0=700\right)$ from the scaled heterogeneous effect model for $M=4$ phenotypes (A–C) and $M=10$ phenotypes (D–F) for various numbers of a subset of phenotypes associated with SNP $\left(M_0\right)$. Three working correlation matrix structures for the joint outcome analyses were considered: I, independent; E, exchangeable; and U, unstructured. Power results were nearly identical with the use of I, E, and U for the 4-DF GEE tests, whereas power results were nearly identical with I and E, but not U, in the 1-DF SMAT; thus, only the results for GEE (U) and SMAT (E and U) are included.