Analyzing Mixed Paired/Two-sample Data

Scott Richter

University of North Carolina at Greensboro


Date: Wednesday, January 25, 2023
Time: 4:00 pm - 5:00 pm
Location: Petty 150

Comparison of two groups is a very common research goal. Such studies may be designed using two independent samples or using matched-pairs data (two dependent samples). Common analysis tools are pooled t-methods or Wilcoxon rank-sum test for independent samples and paired-t methods or Wilcoxonsigned-rank test for paired samples. However, in some cases the data has characteristics of both independent and paired samples. For example, in a matched-pairs study only one observation is obtained for certain pairs.Similarly, some experiments are designed in which both treatments are applied to certain units, while other units receive only one treatment. In each of these examples the resulting data consists of both paired and unpaired data.Several methods that allow the use of the paired and unpaired data in a single analysis have been proposed. I will introduce these methods and discuss some recent extensions, including several student projects.