[Skip to Navigation]
December 1968

The Power of IMPS in Psychiatric Drug Research: As a Function of Sample Size, Number of Raters, and Choice of Treatment Comparison

Author Affiliations

Washington, DC
From the Biometric Laboratory, The George Washington University, Washington, DC.

Arch Gen Psychiatry. 1968;19(6):689-699. doi:10.1001/archpsyc.1968.01740120049008

IN A recent publication, Overall et al1 point out that studies concerned with the evaluation of psychiatric drug treatments have exhibited a certain restricted diligence in regard to their data analyses. Attention in these studies has been principally directed toward the reduction of a particular type of statistical error, ie, of erroneously concluding that treatment differences exist, when in fact they do not. This diligence has not been adequately extended, however, to other considerations in analysis. Attention to the converse type of error, that of concluding that treatments are equivalent when in fact a true difference does exist, has been greatly neglected. Failure to attend sufficiently to this latter consideration has in the past and will continue to result in failure to detect genuine and substantial treatment effects in certain cases where they do exist.

The power of a statistical test is defined by Winer in symbolic fashion

Add or change institution