Bruce Weaver wrote: > Here are the two cents I wanted to throw into the discussion: Measures of > effect size are not particularly useful in certain areas of research where
I'd add further, that tiny effects (in terms of proportion of variance) might be behaviourally important. I remember reading an article that showed that people were sensitive to tiny (millisecond costs) in their allocation of resources on a computer-based task. So such tiny effects might also have surprising practical importance. I'm unconvinced that standardized (and probably unstandardized) effect sizes are always (or even often) useful measures of practical significance/importance. In fact I'd argue that requiring, say, a fixed effect size to publish a study would be far more damaging than relying on fixed significance levels. The main thing is to get over the idea that there are other things to pay attention to than just p values, standardized effect sizes or whatever. Thom . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
