The purpose of the current study was to investigate the effectiveness of four methods of handling missing data. Effectiveness was defined as the probability of reproducing the covariance matrix of the target sample accurately. Effectiveness of the missing data methods was assessed by manipulating the proportion of cases containing missing values and the sample size. The missing data methods studied were: (1) listwise deletion; (2) pairwise deletion; (3) regression; and (4) expectation maximization. The initial sample contained data from a study of 4,645 high school students who reported working while in school. This study replicated E. Witta's (2000) study with an addition to the sampling procedure. The pattern of missing values was used to create missing values in complete cases and compared to the case replacement method. Thus, the current study was also testing the effects of the case replacement sampling procedure used by Witta and the variable value replacement method used in this study. Results indicate there are differences in decisions concerning effectiveness of a missing data method due to the way the sample was created and the proportion of missing values used. (Contains 4 tables and 19 references.) (SLD)