TY - JOUR

T1 - Double quantile regression accurately assesses distance to boundary trade-offs

AU - Cardoso, Gonçalo C.

PY - 2019

Y1 - 2019

N2 - Boundary trade-offs are common among ecological, life-history, behavioural and other traits. Depending on the traits studied, distances of data points to boundary trade-offs can indicate ecological or life-history strategies, or behavioural performance. Quantile regression tests the statistical significance of boundary trade-offs, but it is unknown whether it provides meaningful benchmarks for evaluating distances to the true trade-offs shaping the data. This is especially relevant when traits limit each other mutually, rather than one independent trait limiting another dependent trait. I used empirical and simulated data to evaluate how quantile regression assesses distance to boundary trade-offs. First, I reanalysed empirical datasets showing upper-bound trade-offs between acoustic traits, which is a field where distances to trade-offs are often used to infer behavioural performance. Second, I simulated data under different assumptions of how boundaries influence density distributions, to test the accuracy of assessing distance to the true trade-offs generating the data. Quantile regression assessed distance to upper-bound trade-offs incongruently in most empirical datasets, strongly influenced by arbitrary decisions on which trait to use as dependent. Simulated data showed that a double quantile regression approach — the consensus of two reciprocal quantile regressions — accurately and robustly assesses distance to the true trade-offs generating the data. The method was robust to low sample sizes and to different assumptions on how boundary trade-offs influence the density distribution of data. Double quantile regression can assess distances to the boundary trade-offs observed in various branches of ecology, from functional and behavioural ecology, to population and macro-ecology.

AB - Boundary trade-offs are common among ecological, life-history, behavioural and other traits. Depending on the traits studied, distances of data points to boundary trade-offs can indicate ecological or life-history strategies, or behavioural performance. Quantile regression tests the statistical significance of boundary trade-offs, but it is unknown whether it provides meaningful benchmarks for evaluating distances to the true trade-offs shaping the data. This is especially relevant when traits limit each other mutually, rather than one independent trait limiting another dependent trait. I used empirical and simulated data to evaluate how quantile regression assesses distance to boundary trade-offs. First, I reanalysed empirical datasets showing upper-bound trade-offs between acoustic traits, which is a field where distances to trade-offs are often used to infer behavioural performance. Second, I simulated data under different assumptions of how boundaries influence density distributions, to test the accuracy of assessing distance to the true trade-offs generating the data. Quantile regression assessed distance to upper-bound trade-offs incongruently in most empirical datasets, strongly influenced by arbitrary decisions on which trait to use as dependent. Simulated data showed that a double quantile regression approach — the consensus of two reciprocal quantile regressions — accurately and robustly assesses distance to the true trade-offs generating the data. The method was robust to low sample sizes and to different assumptions on how boundary trade-offs influence the density distribution of data. Double quantile regression can assess distances to the boundary trade-offs observed in various branches of ecology, from functional and behavioural ecology, to population and macro-ecology.

KW - behavioural performance

KW - boundary limits

KW - boundary trade-offs

KW - double quantile regression

KW - quantile regression

KW - statistical methods

U2 - 10.1111/2041-210X.13193

DO - 10.1111/2041-210X.13193

M3 - Journal article

AN - SCOPUS:85066066264

VL - 10

SP - 1322

EP - 1331

JO - Methods in Ecology and Evolution

JF - Methods in Ecology and Evolution

SN - 2041-210X

IS - 8

ER -