Fuzzy Inference in the Analysis of Non-interval Data

An inference engine using fuzzy logic is proposed for the analysis of Likert-type questionnaire. This method was used to understand and incorporate the imprecision of items in a questionnaire so that a single score that encompassed the different scales of the questionnaire could be created. A parent-rated questionnaire called the Parent Checklist of Peer Relationships (PCPR) was used as an example. A single Fuzzy Inference score was calculated that accounted for both prosocial and aggressive behaviors. This score was significantly correlated with the PCPR scale scores, suggesting that the Fuzzy Inference score is valid. The Fuzzy Inference score and the PCPR scores were correlated with independent measures of behaviors based on previous coders’ behavioral ratings. The Fuzzy Inference score was considered to be a better correlate of the earlier behavioral scores because it yielded significant correlations whereas the PCPR score correlations appeared erratic.