A Method for Imputing Ordinal Responses at the Intersection of Misfitting Items and Persons in a Guttman Scale

This study introduces a novel method for imputing misfitting ordinal responses at the intersection of misfit persons and items within a Guttman scale. The imputation process involves replacing misfitting responses with the nearest integer to the average of surrounding responses. Application of this technique to a university depression survey dataset led to reduced mean square fit statistics for misfitting items. Additionally, compared to removing misfit persons, this method improved homogeneity of estimates in both item difficulty and fit statistics. This imputation approach offers an effective alternative to traditional row-wise or column-wise deletions, which often result in substantial data loss.