Premium
Distinguishing between Lévy walks and strong alternative models: comment
Author(s) -
Auger-Méthé Marie,
Plank Michael J.,
Codling Edward A.
Publication year - 2014
Publication title -
ecology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.144
H-Index - 294
eISSN - 1939-9170
pISSN - 0012-9658
DOI - 10.1890/13-1484.1
Subject(s) - content (measure theory) , ecology , computer science , econometrics , geography , mathematics , biology , mathematical analysis
Reynolds (2012) proposed that power spectra and the first-significant-digit law of Benford (1938) can be used to distinguish between movement data arising from a Levy walk (LW) and data from a strong alternative model: the composite correlated random walk (CCRW). Under the CCRW model, the animal typically switches between two behavioral phases, one with a relatively large mean step length, representing movement between patches of food for example, and one with a smaller mean step length, representing movement within a patch (Benhamou 2007). These are sometimes referred to as extensive and intensive movement phases (Knell and Codling 2012). The results of Reynolds (2012) are important as CCRWs can produce movement patterns that appear similar to LWs, even though the underlying processes are quite different (Benhamou 2007, Plank and Codling 2009, Codling and Plank 2011, Gautestad 2012). In this comment, we show that the results of Reynolds (2012) are not robust to changes in the parameters of the CCRW model and that the proposed methods cannot always reliably distinguish between different movement models. We further discuss a number of methodological points related to the study of Reynolds (2012) that should be considered when testing movement models. Reynolds (2012) used simulations to generate step lengths for LWs and for CCRWs. Similarly to others (Plank and Codling 2009, Codling and Plank 2011), he showed that simulated CCRW data can be misclassified as arising from a power law (i.e., as being from a LW) using standard comparative tests based on Akaike weights (Edwards 2008). In addition, Reynolds (2012) claimed that a test of absolute fit (G test), which has been suggested as a method of resolving this issue (Auger-Methe et al. 2011), does not correct this misidentification. Reynolds (2012) proposed two alternative approaches for distinguishing between a LW and a CCRW, based on (1) the power spectrum of the time series of turning points in the movement path and (2) Benford’s law of first significant digits.