The characteristics of missing data in statistical analysis are outlined, and techniques to deal with missing data are explored using a real data set that contained pretest and posttest measures of kindergarten children. In the first part of the investigation, randomly created missing values for 5%, 10%, 20%, and 25% of the sample were supplied for a sample that originally had no missing values. Four missing data techniques (listwise deletion, mean substitution, adjustment-cell mean imputation, and regression imputation) were then implemented, and the results were compared with those of analysis of variance tests. In a second part of the study, the missing data techniques were applied to a real missing data problem, using the same subsample (443 kindergarten students) with pretest data missing for 83 students. Results indicated that disparate results may be obtained with various missing data techniques. When faced with missing data, researchers should investigate whether the data are missing at random or for an identifiable reason, apply some missing data techniques, and consider the consequences carefully if different techniques lead to dissimilar conclusions. (Contains 1 figure, 12 tables, and 18 references.) (SLD)
