Computational protocol: Can Nocturnal Flight Calls of the Migrating Songbird, American Redstart, Encode Sexual Dimorphism and Individual Identity?

Similar protocols

Protocol publication

[…] We applied a permutational multivariate analyses of variance (pMANOVA, described by Anderson []) to the remaining feature measurements after the correlation analysis. Individual identity was nested while age, gender, and variant type were applied as fixed factors to ensured that calls from the same bird would not be compared to one another when examining levels of similarity among and within factor groups. Significantly high or low similarity within calls from the same bird did not influence the results of this analysis []. We calculated a pMANOVA using 999 iterations with the adonis function in the vegan package (2.3–1, []).To illustrate differences between significantly different qualitative factors, we calculated pairwise similarity scores among all flight calls in our dataset by applying a random forest decision tree (after []; randomForest 4.6–12 [], R version 3.2.2 [], ntree = 4999) to the feature measurements calculated for each flight call using the default settings. These scores represented the degree of similarity for every call pair, resulting in a 180 x 180 similarity matrix that identified each call’s proximity to its neighbors []. Additional random forest analyses were run using similar parameters, only supervised by significant fixed factors. A separate supervised random forest was run per significant fixed factor. Therefore, from the resulting matrices ordination Principal Coordinate Analysis (PCoA) plots of the first two principal coordinates groupings could be generated. This translated the multivariate dispersion relationships within our dataset into a collection of points plotted in a two-dimensional coordinate system.To determine whether average call similarity was higher among calls from the same individual, we calculated mean similarity among the five calls from a single bird by averaging the 10 pairwise similarity values for these calls, yielding within-individual mean call similarity scores for each 36 individuals. Next, we calculated mean similarity among the five calls from a single individual to all other calls in the dataset (i.e., averaging 5 x 5 x 35 = 875 similarity values from the similarity matrix), yielding 36 between-individual mean similarity scores. We then applied a two-sample t-test to determine whether within-individual mean call similarity was significantly different from between-individual mean call similarity per individual. To visually represent these similarities, we chose two-dimensional non-metric multidimensional scaling (NMDS) plot. Each data point in this plot represents a single flight call, and the distance between every pair of points corresponds to the pairwise similarities of those calls. […]

Pipeline specifications

Software tools vegan, randomforest
Application Miscellaneous