Factors Affecting the Analysis of Growth Patterns of Large Mammals

Publisher Website
Google Scholar
Abstract
Growth of large mammals under natural conditions often is studied by regressing the mass of collected individuals against their estimated ages; however, this practice might lead to biased estimates of growth. Monte Carlo simulations were used to assess effects of nonuniform sample distribution and of imprecise and erroneous age determination on estimates of growth for male white-tailed deer (Odocoileus virginianus) using the Richards sigmoid growth equation. Effects of errors in age determination and sample distribution also were evaluated over a range of values of growth parameters, and biases were compared between estimates obtained from Richards, Gompertz, and von Bertalanffy equations. Variation in birth dates of individuals tended to increase the variance, but not the bias of growth estimates. Assigning individuals to the wrong year class resulted in biased estimates of all three growth parameters, with the shape parameter m being affected most. This bias was most pronounced when the sampling distribution was nonuniform and was not consistent for growth curves of different shapes. It appears that little confidence should be placed in estimates of m from data that may contain animals assigned incorrect ages. Asymptotic mass (W∞) and growth period (T) were relatively insensitive to this effect; however, a nonuniform sample distribution increased the bias of these estimates. For our data, estimates of W∞ and T obtained from the Richards equation were less biased than estimates obtained from Gompertz or von Bertalanffy equations.