Analysis of variance of population genetic parameters

Two-Factor Analysis of Variance was performed using General Linear Model Multivariate tests (SPSS) to investigate whether the factors “year” (2010 versus 2011) and/or “condition” (unaffected versus oiled populations) influenced the dependent variables measured to determine population and mating structure. Significant deviations from equal error variances were detected in the number of offpring per parent for the factor ‘condition’ (Levene’s P = 0.03, df = 1,11, F = 6.18) and with marginal significance when all factors and combinations were considered (P = 0.10, df = 3,9, F = 2.79). Log transformation removed the issue (P > 0.20). Unequal variances were also detected in the percentage of fullsib pairs (Levene’s test: P = 0.001, df = 3,9, F = 14.85) when all factors and combinations were considered, but not for ‘condition’ (P = 0.15, df = 1,11, F = 2.48). Transformations failed to remove unequal error variance in this dependent variable. Thus, an additional independent samples t-test not assuming equal variances was performed.