Principles of defining reliable change indices

In this article, several salient measures for determining reliable change are scrutinized. The classic null hypothesis method is compared with more recent procedures based on interval estimation of the true change, including Kelley's formula. The latter category of methods are shown to entail serious drawbacks. If Kelley's formula is expanded to a null hypothesis method (including a correct treatment of the stochastic character of the sample information), the classic method reveals itself as a large sample approximation. We conclude that the classic method is undeservedly regarded inferior by the authors who proposed new indices.