Measuring magnitude of treatment effects in agricultural experiments

The estimation of treatment effect is an aspect of data analysis in agricultural experiments that is neglected. Such an effect size value cannot be determined through the usual Null Hypothesis Significant Testing (NHST) in agricultural research. Thus, the current study aimed to describe the foundation supporting the estimation and reporting of effect size. In the re-analysis of two unpublished data sets concerning the effects of four feeding treatments of Madre de Agua (Trichanthera gigantea) on the average final weight of pigs and on the feed cost per pig, effect sizes of treatments were measured following the omega squared algorithm procedure. This algorithm is based on the ratio of variances, which result is a unit less measure on the magnitude of treatment effect. Results of re-analysis successfully estimated the magnitude of effect equivalent to ω2=.08 or 8 % for the average final weight of pigs and ω2 = .9911 or 99.11 % for the feed cost per pig. These treatment effects are considered large based on the previously reported standards, and therefore substantial and appropriate recommendation may be given. The importance, implications and practical application of this effect size measure are discussed.