Sensitivity of an experiment (Chapters 7 and 8.1)

Power of an experiment: 1-b: ability to recognize treatment effects

Increase of power through:

• reduction of treatment variability
• analysis of control factors instead of randomizing them
• reduction of subject variability (or within-subjects designs or analysis of co-variance)

Effect size (treatment magnitude): Measure for the strenght of a treatment effect

         variations among the experimental treatment means
w2 = ---------------------------------------------------------------------
         (variations among the experimental treatment means +
         variability among the treatment groups)

w2 is independent of the sample size; it is a measure of the proportion of the total variation that is explained (accounted for) by the treatment.

"Small" /  "medium" / "large" effects in the behavioral sciences: w2 = .01/ .06 / >.15

The following factors are dependent on each other in the following way:

           Power * Effect size * sample size * significance level = quality of experiment

where * means something like "positive contribution"