Data, Sample Sizes and Statistics Affect the Recognition of Species of Bats by Their Echolocation Calls

摘要: Identification of bat species based on analysis echolocation calls can be affected by the way data are manipulated, diversity species, and call variability. We document effects sample sizes a priori assignment outcome discriminant function (DFA) multinomial logistic regression (MLR) features calls, determine which most useful for identification. used recorded eight readily distinguishable features, including molossids, emballonurids moormopid at sites in Belize, Brazil, Mexico. On individual we measured four features: frequency with energy, highest lowest frequencies durations obtained from sequences consisting 10 calls. Cluster multiple analyses variance indicated significant differences between different species. Outcomes DFA MLR were both (numbers numbers sequences) subjective approach that researchers take to their (i.e., categorizing or species). Levels variation some our often precluded use single making call-based identifications. Accurate documentation variability behavior sympatric bats is prerequisite an effective sound-based survey.