Effects of forest size and selective logging on species richness and community composition of six animal taxa in Kakamega Forest.

Community composition was quantified from non-metric multi-dimensional scaling (NMDS) (site scores on first axis). Given are regression parameters of minimal adequate models. Human disturbance increases from low to high values (forest size was multiplied by −1). As effect directions in NMDS are arbitrary, effects on community composition are given as positive values. Species richness of leaf litter fauna (no species data) and army ants (only two species) were not tested. Spatial autocorrelation was assessed by Moran's I values, derived from the model residuals and a spatial weights matrix from the four nearest neighbours of each BDO.