Method of distinguishing defective responses from a questionnaire data on the basis of internal consistency of responses.

Applying to semantic differential data

Abstract

In this paper, we proposed a method to distinguish defective responses from a questionnaire data on the basis of internal consistency of responses. In general, several questions in a questionnaire relate each other, so higher correlation coefficients are observed among them. However in the case of defective responses, the coefficients may be lower because of lack of internal consistency among responses. In the method proposed, we focus on correlation coefficients calculated for each respondent and compare them among respondents so as to identify defective respondents from normal ones. Applying our method to semantic differential data, we could appropriately distinguish several respondents considered to be defective.