184 Note: Loss of Power through Ineffective Pairing of Observations in Small Two-Treatment All-or-None Experiments

Publisher Website
Google Scholar
Abstract
Matching of cases which, unknown to the experimenter, bears no relation to the probabilities of response under either treatment is shown to be disadvantageous for small sample sizes. The disadvantages of pairing, in terms of reduction in the ability to detect experimental differences, which is known to be essentially absent for large samples, is demonstrated to vanish for sample sizes as large as 30 pairs.

This publication has 0 references indexed in Scilit: