Tests for Two Related Samples: Pretest-Posttest Measures for a Single Sample

Abstract

In Chapter 4, we examined nonparametric tests that could be used to determine the extent to which a single sample is similar to a hypothesized theoretical sample. In health care research, we frequently try to assess whether a particular intervention is effective with a certain population. In our hypothetical example from Chapters 1 and 3, we were interested in evaluating the effects of a staff-initiated intervention to reduce the number of sleep environment interruptions for hospitalized pediatric cancer patients on the frequency of their nocturnal awakenings, levels of fatigue, and distress compared to a usualcare group. In this study, there may be certain characteristics of the sample, such as type of cancer, age, or gender of the child, that are known from prior research to confound and potentially misrepresent the outcomes of the intervention. Two approaches that can be used to address this problem prior to the intervention are the following: (1) matching the subjects with regard to these extraneous confounding variables and then randomly assigning one of the pairs to the control and the other to the intervention group and (2) using each subject as his or her own control. Data being analyzed then become paired, either through the use of related samples or through repeated observations on a single sample. A third occurrence of matched pairs can occur when the researcher has sampled observations in pairs (e.g., husband and wife). Although each member of the pair may have separate scores on a dependent variable (e.g., marital satisfaction), there is reason to believe that knowing the scores of one member of the pair (e.g., wife’s marital satisfaction) will give information about the scores of the other member (husband’s marital satisfaction).