Schematic representation of the simulation settings and results.

<p>Schematic representation of the simulation settings (a–f) and radar plot (g) of the power to detect 1 genetic variant (GV) explaining .5% of the phenotypic variance in 12 simulation settings. The power radar plot (power running from 0 (midpoint) to 1 (outer edge)) displays the power for the univariate sum score analyses (blue), MANOVA (green), and TATES (red). The phenotypic correlation structure was either due to one common factor (a,e), multiple underlying latent factors (b,c,d), or a network model (f). Within these phenotypic settings, the GV either affected multiple phenotypes via a common factor (a,b,c), or affect a single phenotype directly (d,e,f). Power results for 12 simulation settings and a GV explaining .5% of the variance are highlighted (g, colour labels corresponds to colour simulation settings; see <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s002" target="_blank">Tables S1</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s003" target="_blank">S2</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s004" target="_blank">S3</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s005" target="_blank">S4</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s006" target="_blank">S5</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s007" target="_blank">S6</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s008" target="_blank">S7</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s009" target="_blank">S8</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s010" target="_blank">S9</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s011" target="_blank">S10</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s012" target="_blank">S11</a>, <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003235#pgen.1003235.s013" target="_blank">S12</a> for more GV effect sizes). Specifically, gA1–3: 1-factor models with GV effect on the factor. gA1: mix of dichotomous, ordinal and continuous phenotypes correlating .36 to .81; gA2: continuous phenotypes correlating .56; gA3: continuous phenotypes correlating .12. gE1–3: 1-factor models with GV effect specific to 1 phenotype. gE1: phenotypes correlate .56 (like gA2); gE2: phenotypes correlate .30; gE3, phenotypes correlate .12 (like gA3). gF1–F3: network models with GV effect specific to 1 phenotype. gF1: phenotypes correlate .56 (like gA2 and gE1); gF2: phenotypes correlate .12 (like gA3 and gE3); gF3: 4 clusters of phenotypes that within clusters correlate .55, and between clusters correlate .13. gC1: 2-factor model, 10 phenotypes per factor, correlating .36–.81 within factors, and a factorial correlation of .5. GV affects only the 2^nd factor. gB1: 4-factor model, 5 phenotypes per factor, correlating .81 within factors, and factorial correlations of .1. GV affects only the 4^th factor. gD1: like gB1 but GV effect specific to 1 phenotype.