I finished up my independent project over the weekend. I ended up going back to my original idea - using Mantel tests to test for niche complementarity in understory flycatchers at La Selva. The data was collected primarily by my advisor, Tom Sherry, in the 1970s with some recent observations collected in 2000 and 2003. The data included habitats, foraging heights, diet (order/family of arthropods consumed, as determined from gut contents), and prey size (again from gut contents) of 15 species. The hypothesis was that birds that were similar in one niche dimension - e.g., habitat - would differ in another niche dimension - e.g., diet - or, at the least, not be similar in more than one.
I took the compiled data, imported it into R, and calculated dissimilarity values for all pairs of species (e.g., 1 vs 2, 1 vs 3, 2 vs 3, and so on...) for each niche parameter. These values were then compiled into matrices, which were then run through Mantel tests. Mantel tests allow you to test for correlations between two matrices - essentially they're like Pearson's correlation coefficients on steroids. Each pair of matrices (diet and habitat, diet and prey size, etc.) was tested for correlations.
As it turned out, there was no correlation between either food-related parameter (diet or prey size) and any other parameter. Oddly enough, though, I found a significant and positive correlation between habitat and foraging height. That is, species that overlap in habitat also overlap in foraging height, and conversely species that occupy different habitats have different foraging heights. Rather counterintuitive, contrasting with both theory and our hypotheses. However, this was a preliminary study, and we're in the process of collecting additional data and obtaining some old data that is currently in another lab, so perhaps this will help. It is true that the gut content sample size is very low - but LSU has an extensive collection, so perhaps we can improve power by boosting the sample size (now there's an idea - but for later).