Analysis of Variety Trials in the Presence of Non-Random Time Effect

The assumption of random year effect in the analysis of variety trials data over locations and years clearly contradicts the practice of conducting the trials over only a few consecutive years. A method of analysis of such trials which assumes that the year-dependent effects are serially correlated is proposed in this paper to cater for the non-randomness in the selection of years of the experiments. A simple autoregressive model was assumed for the serial relationship. A reduction in the variance estimate of the difference between two variety means was obtained using this method.