Variation in community structure is mediated by interactions between species traits and a site's environmental characteristics. Previously, data on community composition at sites has been employed to correlate trait and environmental variables (e.g., RLQ analysis) and to predict community-level expression of quantitative traits (i.e., community aggregated traits). Here, we demonstrate that the selection ratio, a method originating in animal resource selection studies, can estimate the ecological filters that site conditions impose on species traits by combining observed community aggregated traits with null models of species availability. This flexible, nonparametric approach expresses the filter at each site as a probability density function for the selection of individuals possessing a given trait value. By doing so, it generalizes the community aggregated trait concept to include categorical as well as continuous traits and allows for both intraspecific variation in trait expression and differences in species availability among sites. The resulting site-level filter functions can be related to environmental covariates by standard statistical approaches (e.g., regression). The method complements existing techniques for analyzing trait-environment interactions in community ecology.