[R] pwr.prop.test and continuity correction

[Thomas Lumley]

On Tue, 16 Dec 2008, Peter Dalgaard wrote:
>
> power.prop.test (sic) is relying heavily on asymptotic normality, as do similar 
> formulas. It doesn't use continuity correction, but if you're working with such 
> small group sizes, I suspect that the correction term is the least of your 
> worries and that direct simulation would be better.
>

In fact, for tests in 2x2 tables, it is fairly straightforward and fast to compute the entire sampling distribution explicitly, over a grid of parameter values.

This gives the exact power (under alternatives) and the exact Type I error (under null). You can also compare different tests and see how much the continuity correction moves the actual Type I error rate away from the nominal rate.

        -thomas

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle