DETECTING THE ORDER OF POPULATION DYNAMICS FROM TIME SERIES: NONLINEARITY CAUSES SPURIOUS DIAGNOSIS

Abstract

Partial autocorrelation and partial rate correlation functions are frequently used to detect the order of the endogenous process generating an observed population time series. Here we uncover a problem with this approach: the diagnosis of spurious second order autocorrelation due to strong nonlinearity in a first order endogenous process, as exemplified by time series data from a population of Soay sheep. Causes and a possible solution are discussed.