Experimental power: The other side of the coin

The article by Mitchell et al. in this issue of Pediatrics is laudable in many respects, including the rigor of its methodology. In a paragraph on page 974, the reader will find one particularly valuable nugget. There the authors write, "...clinically important differences... were unlikely to have been missed. Specifically, for GCI (general cognitive index) a difference of 15 points would have been found 91% of the time, and a difference of 20 points 99% of the time, while a difference of 10 points would be found 64% of the time."  That is pure gold. Not only do the authors tell us that their study failed to find a certain effect (a difference in GCI between cases and comparisands), but they then tell us how likely it was, given the size and design of their study, that they missed an effect which we might care about. (May 1980)