A Monte Carlo investigation of three estimates of treatment magnitude: Epsilon squared, eta squared, and omega squared.

The effectiveness of three estimators of treatment magnitude were compared numerically for samples from a normal and exponential distribution. The estimators were compared for Cohen's (1969) definitions of small, medium, and large population treatment effects. It was found that omega squared is a more accurate estimator while eta squared has the smallest sampling variability. RESUME Comparaison numerique de I'efficacite relative de trois techniques d'extinction de la grandeur d'un effet experimental, pour des echantillons distribues de facon normale ou exponentielle . La comparaison utilise les definitions donnees par Cohen (1969) des effets experimentaux chez des populations restreintes, moyennes et etendues. L'etude montre que I'omega carre est la plus exacte des techniques d'estimation et que I'eta, carre a la plus petite variability echantillonnale.