Eliminating invalid self-report survey data.

A sample of 6,370 students in Grades 6 to 8 completed a questionnaire on their attitudes and use of alcohol, tobacco, and other drugs. A subsample showed questionable data based on three criteria: missing responses, invalid responses, and inconsistent responses. Analysis indicated that this subsample was significantly different from the main group on demographic variables and self-reported life-time tobacco use. Results support efforts to identify and eliminate invalid data.