Analysis of a Multivariate Response Resulting from Observations Taken Over Time on Experimental Units in a Randomized Complete Block Design

Abstract

In assessing the effect of any treatment to control pests, the effect of the treatment over time is of as much interest as the effect immediately following treatment. Experiments designed to compare the effect of several treatments use time and experimental material more efficiently by making several observations over time on each experimental unit. In studies conducted in this manner, correlations between observations taken on the same experimental unit may complicate the analysis and may make consideration of time as a factor of interest in the analysis more difficult. In this article, advantages and disadvantages of several approaches for analyzing multivariate response data resulting from observations taken over time on experimental units in a randomized complete block design are discussed. Profile analysis is a multivariate technique that is restricted by fewer assumptions than many other techniques. Profile analysis of such multivariate response data is described in detail. This technique can be used, in the context of analyzing pest management data, to compare trends in response over time with various treatments. An example is presented to illustrate the method.