Skip to main content
. 2021 Nov 1;113(1):79–90. doi: 10.1093/jhered/esab056

Figure 4.

Figure 4.

Violin plots showing magnitude of ψ estimates for on-diagonals (reciprocal IGEs involving the same trait) versus off-diagonals (IGEs acting on different traits). Points show the 225 estimates of ψ analyzed, with significance in the original study indicated with red circles (significant), green triangles (nonsignificant), or blue squares (not reported). The black points and bars show the fitted values and standard error obtained with a linear mixed model including ψ as the response variable, IGE type (on-diagonal or not) as a fixed effect and study identity as random factor. Note that all values where |ψ|>1 were removed for this analysis, as it is impossible to estimate ψ(corr) for these. If these values (n = 13) are included in the analysis uncorrected, the outcome is qualitatively the same: the magnitude of reciprocal IGEs with feedback is significantly greater. See the main text for statistical details.