Abstract
A study is made of certain designs of Latin square type, namely, Latin squares with (a) a column added, (b) a row and a column added, and (c) a column added and a row omitted. Recommendations are given for the analysis of data by a method leading to an unbiased test. It is shown that sufficient randomization is obtained by writing down a possible configuration of plots and permuting rows and columns at random.

This publication has 4 references indexed in Scilit: