Analysis of Short-term Selection Experiments: 1. Least-squares Approaches

Abstract

This chapter examines short-term (a few generations) selection response in the mean of a trait. Traditionally, such experiments are analyzed using least-squares (LS) approaches. While ordinary LS (OLS) is often used, genetic drift causes the residual to be both correlated and heteroscedastic, resulting in the sampling variances given by OLS being too small. This chapter details the appropriate general LS (GLS) approaches to properly account for this residual error structure. It also reviews some of the common features observed in short-term selection experiments and examines experimental designs, such as the use of a control population versus a divergence-selection approach. It concludes by discussing another linear model used mainly by plant breeders, generation-means analysis (GMA), wherein remnant seed for several generations of response are crossed and then grown in a common garden. Such an analysis can provide insight into the genetic nature of any response.