ABSTRACT
A method is described that incorporates covariate information into the analysis of nonreplicated experiments with factorial and fractional factorial treatment structures. The data are used to estimate the slope corresponding to the covariate (or slopes for multiple covariates) and then the estimates of the effects are adjusted for differing values of the covariate or covariates, i.e., the methodology provides estimates of the factorial effects adjusted for the covariate or covariates or after removing the confounding effects of the covariates on the factorial effects. Regression analyses are used to provide estimates of the factorial effects and of the sampling variance. The half-normal probability plot is one method for analyzing nonreplicated experiments without covariates and that method is used to compare the results of the covariance analysis. Five examples are presented to demonstrate the methodology. Three examples involve a single covariate, one example involves two covariates, and one example involves one covariate with unequal slopes for the levels of one of the factors. Because of the limited number of data points in nonreplicated experiments, there is a limit as to the number of covariate parameters that can be estimated.
