Signal or noise? A null model method for evaluating the significance of turnover pulses

Bibliographic Collection: 
APE
Publication Type: Journal Article
Authors: Barr, W. Andrew
Year of Publication: 2017
Journal: Paleobiology
Pagination: 1-11
Date Published: 2017
Publication Language: eng
ISBN Number: 0094-8373
Abstract:

Patterns of turnover in the mammalian fossil record have been interpreted as reflecting “pulses” of originations and extinctions hypothesized to be driven by climate change. However, criteria for determining what constitutes a meaningful pulse have been idiosyncratic, and investigations of turnover patterns in mammals have yielded mixed results. This study presents simple simulations of fossil records in which origination and extinction probabilities for each lineage are held constant. Nonetheless, the total number of turnover events per time bin varies stochastically, producing statistical “noise.” Various simulation and analytical assumptions are examined to determine their impact on the type I error rate (i.e., how often “pulses” are detected in a purely stochastic process). Results suggest that simple analytical parameters (length of time bins and turnover-pulse criterion) have predictable and straightforward effects on false-positive rates. Furthermore, “pulses” of turnover of a magnitude similar to that observed in the terrestrial mammalian fossil record may be quite common under realistic analytical conditions. The null turnover model offers a practical way to evaluate the significance of observed turnover events in future empirical studies of the fossil record. In evaluating the significance of a “pulse” of fossil origination or extinction events, analytical parameters can be explored using this null model to determine the approximate type I error rate for a set of parameters. Because false-positive rates are shown to be quite high, functional trait-based approaches may offer more reliable indicators of the impact of climate change on turnover dynamics.

DOI: https://doi.org/10.1017/pab.2017.21
Export: