MAKİNE ÖĞRENİMİNE DAYALI ÇOKLU ATAMA YÖNTEMLERİNİN PARAMETRİK OLMAYAN ÇOKLU KARŞILAŞTIRMALARA UYGULANMASI VE BENZETİM ÇALIŞMASI

Abstract

Longitudinal data, in which repeated measurements are made on the same subjects, eliminate potential differences among the subjects. In longitudinal data, missing data can occur by design or completely random. The Skillings-Mack test is used instead of the Friedman test for longitudinal data with missing observations that are non-normally distributed. Nonparametric multiple comparisons need to be performed if a significant difference exists among groups. In this study, we propose a new approach by applying four methods to nonparametric multiple comparisons of longitudinal data that are non-normally distributed. The four methods are two nonparametric multiple imputation methods based on machine learning, one nonparametric imputation method (random hot deck imputation), and the listwise deletion method. We assume two missing data mechanisms. After implementing different scenarios in a simulation study, the listwise deletion method is inferior to the other methods. The two nonparametric multiple imputation methods are superior to the other methods for moderate and small sample sizes with well-controlled type 1 error. Therefore, we propose the two multiple imputation methods for nonparametric multiple comparisons of longitudinal data with missing observations. Moreover, the proposed approach was also applied on a real data set. In this example, the proposed multiple imputation methods yielded results similar to those of the real dataset without missing observations at the beginning.