Statistical Analysis of Line‐Source Sprinkler Experiments and Other Nonrandomized Experiments Using Multivariate Methods

A recent paper by Hanks et al. (1980) suggested that there was no statistical test available for comparing the effects of irrigation level on yield because irrigation levels are often applied systematically without randomization. This paper shows how multivariate methods can be used to obtain an appropriate statistical test. The method can also be used for many other nonrandomized experiments. One such experiment occurs when readings on the same experimental unit are recorded at several different time periods. Another occurs when measurements are made on several different parts of the same plant, e.g. stem, leaf, etc. The methods provide tests for equal cultivar effects; irrigation effects, and also cultivar × irrigation level interaction.