Parametric and nonparametric methods used in analyzing partially correlated data and simulation study

Partially correlated data is a situation that missing values appear in a dependent data for two groups. In dependent grouped data, repeated measures are made on the same individuals. Partially correlated data consists of two subgroups. These are paired data and unpaired data. In paired data, no missing data is observed. In unpaired data, missing data occurs in one of two groups. Partially correlated data is a combination of paired data and unpaired data. As a result of the literature review, eight parametric and eleven nonparametric methods were found to analyze the partially correlated data. The aim of this study is to compare the methods introduced in the study with traditional methods and to determine the method that performs the best. In this study, two applications are given for parametric and nonparametric methods and listwise deletion method selected as conventional method were compared with the methods introduced in the study. Furthermore, two simulation studies were carried out. In the simulation study on parametric methods, 10%, 30% and 50% of the whole data set were deleted at random and listwise deletion method was taken as a conventional method. As the result of the simulation study, it was determined that four of the eight methods were superior to the listwise deletion method and the most powerful method was weighted t test. In the simulation study on nonparametric methods, randomly 25% of the whole data set was deleted and listwise deletion method was taken as the conventional method. As the result of the simulation study, it was determined that six of the eleven methods were superior to the listwise deletion method and the most powerful method was the new signed-rank test. Key Words : Paired data, unpaired data, partially correlated data, missing data, missing completely at random