How accurately data predict flight behavior

To assess how well extension ROM, wing shape, or body mass could be used to predict species’ flight behavior group, we used phylogenetic flexible discriminant analyses (16). Here, the dependent variable was the set of nine flight behavior groups. A separate analysis was carried out for extension ROM, wing shape, and body mass datasets. For the analyses involving extension ROM and wing shape, scores from principal components that accounted for 95% of the parent dataset’s variance were used as predictors. For the analysis using body mass, values were natural log–transformed before being used as predictors. Across all analyses, we used two types of sensitivity analyses of prediction performance. First, the maximum clade credibility tree along with 1000 posterior distribution trees was each used to inform the discriminant analysis. Because we found the analysis invariant to phylogeny, only the maximum clade credibility tree was used thereafter. Second, species’ data were jackknifed: In each of 61 iterations, a single species was held out and discriminant axes were informed by data from the other 60 species. Prediction performance was then determined by the proportion of correct entries in each resulting confusion matrix. To determine how well each empirical dataset performed against random chance, we additionally randomized rows within each dataset and reran all discriminant analyses along with phylogenetic sensitivity analyses. Cohen’s κ, a measure of effect size in empirical versus random performance (17), was then computed for each iteration of this sensitivity analysis using the formulaκ=pope1pewhere po was the prediction performance when empirical data were used, and pe was the prediction performance for when randomized data were used.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.