A smoothed residual based goodness-of-fit statistic for nest-survival models

Abstract

Estimating nest success and identifying important factors related to nest-survival rates is an essential goal for many wildlife researchers interested in understanding avian population dynamics. Advances in statistical methods have led to a number of estimation methods and approaches to modeling this problem. Recently developed models allow researchers to include a covariate that varies by individual and time. These techniques improve the realism of the models, but they suffer from a lack of available diagnostic tools to assess their adequacy. The PROC NLMIXED procedure in SAS offers a particularly useful approach to modeling nest survival. This procedure uses Gaussian quadrature to estimate the parameters of a generalized linear mixed model. Using the SAS GLMMIX macro, we extend a goodness-of-fit measure that has demonstrated desirable properties for use in settings where quasi-likelihood estimation is used. The statistic is an unweighted sum of squares of the kernel-smoothed model residuals. We first verify the proposed distribution under the null hypothesis that the model is correctly specified using the new estimation procedure through simulation studies. We then illustrate the use of the statistic through an example analysis of daily nest-survival rates.

Citation

Sturdivant, Rodney X.; Rotella, Jay J.; Russell, Robin E. 2008. A smoothed residual based goodness-of-fit statistic for nest-survival models. Studies in Avian Biology. 34: 45-54.