A procedure for combining menstrual cycle data.

A procedure is presented which overcomes most of the conceptual and statistical problems associated with the combining of data from heterogeneous menstrual cycles for graphical or statistical analyses. This procedure is based upon an initial normalization of the raw data to eliminate extraneous between-cycle variability, followed by the assignment of the data to a set of discrete cycle phases using a weighted-average technique. The efficacy of this procedure is compared to that of seven other published categorization methods by examining the proportion of variance accounted for and the P values from analyses of variance computed for 17 beta-estradiol and olfactory sensitivity measures. A major advantage of the proposed procedure is that it allows for the grouping of data from entire cycles (including menses) on the same figure without exhibiting points from heterogeneous sectors of individual cycles and without changing the sample size as distance from the midcycle LH surge increases. Thus, this procedure provides equal sample sizes for all phases of the cycle, allowing repeated-measures parametric statistical analyses to be performed. Data are presented which suggest that the categorization of menstrual cycle data without carefully established cycle phases can lead to quite erroneous conclusions.