The purpose of this study was to compare the effects of four Guttman-based indices (SCI, MCI, NCI, W&B). Aberant response data were randomly simulated and compared under four conditions (test length, severity of aberrance, types of misfit, and percentage of aberrant respondents). Also modified cutoffs (the 80(superscript th) percentiles) were used to replace the cutoffs originally announced by their various thumb rules. Results showed that (1) significant differences of detection rates existed among four indices (W&B>SCI>MCI>NCI) under original thumb-rule cutoffs, but the rank changed (W&B>NCI>SCI>MCI) under the modified cutoffs. (2) detecting performanes of the four indices were more consistent in modified-cutoff situation than in thumb-rule situation. (3) either using modified or thumb-rule cutoff almost indices were inferenced by the four simulated conditions.
The purpose of this study was to compare the effects of four Guttman-based indices (SCI, MCI, NCI, W&B). Aberant response data were randomly simulated and compared under four conditions (test length, severity of aberrance, types of misfit, and percentage of aberrant respondents). Also modified cutoffs (the 80(superscript th) percentiles) were used to replace the cutoffs originally announced by their various thumb rules. Results showed that (1) significant differences of detection rates existed among four indices (W&B>SCI>MCI>NCI) under original thumb-rule cutoffs, but the rank changed (W&B>NCI>SCI>MCI) under the modified cutoffs. (2) detecting performanes of the four indices were more consistent in modified-cutoff situation than in thumb-rule situation. (3) either using modified or thumb-rule cutoff almost indices were inferenced by the four simulated conditions.