Abstract
The study examined the reliability and validity, including the factor structure, of a 10-item abbreviated version of the Center for Epidemiological Studies–Depression (CES—Depression) scale when administered as part of a mail questionnaire. It also examined patterns of nonresponse to items and the effects of imputation of data for missing items on the factor structure of the inventory. A problem of missing data has been reported even with interview administration of the CES–Depression. Researchers have varied considerably in the amount of imputation used to replace missing datapoints. In this study, factor structures varied when items were imputed. In addition, those subjects with complete data were compared with those with up to two imputed datapoints. Those subjects with imputed data were more likely to be female, have lower functional status scores, lower self-reported health status, more advanced age, and a greater number of depressive symptoms than those with no missing data. While the estimate of coefficient alpha of .78 indicated the inventory was reliable, the effects of missing data on construct validity were problematic.