Research Design and Analysis Clinic

Abstract

The important design implication of the n/cell question is that, whenever possible, one should have at least 2 observations per cell. We should not disregard additional factors in a study simply because they will create small cell sizes. In other words, the credibility of a study should not be questioned because of small cell size, per se. In addition to the greater information obtained from the additional factors and their interactions, their inclusion will tend to increase the power of the other significance tests in the design when any of the null hypotheses associated with the additional factor are not true. Even if all null hypotheses with which the additional factors are associated are true, the power of the other significance tested is reduced trivially, except when their inclusion reduces the degrees of freedom of the error term to perhaps 30 or less. The size of slices into which a pie is cut does not alter the size of the pie.