A Note on Handling Missing Data

Abstract

Researchers are often confronted with missing data for some subjects and variables. The usual method of handling this problem deletes subjects with missing data and thereby reduces the sample size. An alternative approach suggested by Cohen (1968) investigates the randomness of the missing data with respect to other variables and permits the researcher to use the information obtained to further understand the relationships among variables. This note compares three methods of treating missing data with respect to the results obtained, the informational value of each method, the population to which the results can be referred, and statistical power.