John Epling writes:
>How big should a study of a diagnostic test be?
>Retrospectively, I guess confidence intervals would
>address this, but is there a calculation analogous
>to a power analysis that would apply to figure out the
>size of the groups beforehand?
Sample size is not important. Just make sure to ask for enough in the
research grant so that you can buy your consulting statistician a new
computer. Just kidding. You are actually pretty close to having the right
answer.
Power calculations are appropriate only when you have a research hypothesis.
The emphasis in a study of a diagnostic test is estimation. You want
accurate estimates of sensitivity, specificity and/or likelihood ratios.
When the focus is on estimation, you determine the sample size through the
width of the confidence interval. You specify how precise (how narrow) you
want your confidence intervals to be. This determines your sample size.
For sensitivity and specificity, use the standard formulas for a binomial
proportion (which can be found in any introductory statistics book). For a
likelihood ratio, the formulas are a bit more complex, but the same
principle applies.
I hope this helps.
Steve Simon, [log in to unmask], Standard Disclaimer.
STATS - Steve's Attempt to Teach Statistics: http://www.cmh.edu/stats
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|