Data analysis
We calculated the percentages of optimal models sharing the same RM-FC
combination among all the five model selection approaches, and between
different pairs, to check for any general trend in RM-FC combination
among optimal models. We also statistically tested the correlation among
the optimal models chosen by the selection approaches for the model
features, viz., ORs, AUCDIFF, AUCTEST,
and average number of parameters, and also the area of suitable habitat
predicted. We interpreted the strength of correlation based on Mukaka
(2012). We used nonparametric Spearman’s correlation since our variables
were not normally distributed, except for few (Table S2) when we
performed a preliminary analysis. We performed all the statistical
analysis in IBM SPSS Statistics 23 (IBM Corp.).