The effect of missing data rate on internal consistency within different conditions

Missing data is problematical for researchers. As common statistical analysis programs depend on full data sets, researchers need solutions to eliminate missing data problem. Although full data sets are created via data imputation by using several methods in literature, it is known that psychometric features of the scale will be affected. In this study, the change in Cronbach α value for data sets having missing data in different rates, has been analyzed in terms of sample size, test length, distribution manner and scoring method. Full data sets in the study have been created as 20-repetitions under conditions designated in Wingen 3 program, and by applying factor analysis to data sets chosen randomly from 20-repetitions, it has been determined that data are one-dimensional. Values have been deleted randomly in the rates of 5%, 10% and 20% in accordance with missing data mechanism with codes typed in R program, where each full data set is created, and missing data sets, where Cronbach α reliability value will be calculated, have been obtained. This study is simulative as the data are created via simulation method and also relational and aims to determine reliability changes. For different scoring methods, as simple size increases, change in Cronbach α value decreases. As the number of items increases, change in Cronbach α decreases. Even if sample size, test length and distribution manner are different, as the rate of missing data increases, change in Cronbach α value increases.