A Reinforced Randomized Block Design with Correlated Errors

In the analysis of variance the errors are assumed to be independent which may not always be true. In this article, a restricted randomized block design under compound autocorrelated structure is considered. A method of analysis for such a randomized block design has been developed. It provides confidence ellipsoid and confidence interval of a set of estimable functions and an estimable function of treatment parameters, respectively. Analogous to Scheffe's method, the multiple comparison technique of judging all possible treatment contrasts has been derived. In the process, a method of estimation of unknown parameters has been also developed. An example (simulated data) illustrates this approach.