Post hoc tests following analysis of variance.

Review of nursing research literature revealed that post hoc comparisons of mean differences were not consistently reported, although their use was warranted. This paper describes the critical features of post hoc procedures and offers guidelines for selecting one method over another. Particular attention is given to how various procedures control Type 1 error. Selection of post hoc procedures considering the investigator's desire for control of Type 1 and Type 2 error also is presented.