Regression methods for spatially correlated data: An example using beetle attacks in a seed orchard

Abstract

We present a statistical procedure for studying the simultaneous effects of observed covariates and unmeasured spatial variables on responses of interest. The procedure uses regression type analyses that can be used with existing statistical software packages. An example using the rate of twig beetle attacks on Douglas-fir trees in a seed orchard illustrates the problem and the recommended statistical procedures. Use of this procedure revealed spatial patterns in beetle attacks that were not otherwise evident. We also found that the covariates, tree vigor, cone crop, and host genotype (i.e., clone) apparently had significant effects on number of beetle attacks within a model that also included location effects. This procedure should prove useful in other situations where spatial trends are of interest in themselves and are not simply viewed as nuisance parameters to be dealt with by the proper experimental design.