I've never done much with effect size ... I'd like a way to state the
magnitude of an effect with two groups on a binary variable---in this case,
the frequency of behavior problems in children in two Canadian provinces.
For example, 25% use drugs in Ontario, 30% do so in Quebec. I have lots of
power to resolve statistical significance, but what can I say about the
magnitude of the difference that doesn't look dumb?
David Klein
|