What it says! Actually, when you see it, it what is really being done is to compare the mean values of something in two or more different groups, it might be mean CORE-OM scores at referral comparing across say five different referral sources, or across male, female and non-binary gendered clients. The key distinction being made is that the values in each dataset are from the same people (usually people in our realm) as opposed to values from the same person on different occasions: that comparison, say of mean CORE-OM scores at referral, assessment, first session and at termination of therapy is a “within subjects” comparison. (Sorry for the non-PC terminology but I have rarely if ever seen “within participants”) . Where a within-subjects comparison is of only two values per person it is often called a “paired comparison” or a “paired test”.
Details #
This term comes pretty much from the quantitative methodology of “analysis of variance (ANOVA)” which is generally used in the form of a parametric testing of whether the differences between the mean scores in the two or more groups are so different that it is “significantly” unlikely (usually p < .05) that differences as large or larger than seen in the data would have arisen by chance alone had the data per group have arisen by random sampling from populations in which the means were identical. (The usual Null Hypothesis Significance Testing “NHST” paradigm.)
The term also applies to related “non-parametric” tests such as the Wilcoxon/Mann-Whitney and Kruskall-Wallis tests which don’t come out of a model in which it is assumed that population distributions are Gaussian.
Try also #
Analysis of variance (ANOVA)
Gaussian (“Normal”) distribution
Non-parametric tests
Null Hypothesis Significance Testing “NHST”
Paired tests
Parametric tests
Repeated measures tests
Within subjects/participants tests
Chapters #
This comes up, in various ways, in chapters 5, 6, 7 and 8.
Online resources #
None yet.
Dates #
First created 4.ii.24.
