Asaduzzaman Khan1, Chi-Wen Chien, Karl S Bagraith. 1. School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane QLD 4072, Australia. a.khan2@uq.edu.au.
Abstract
OBJECTIVE: To investigate whether using a parametric statistic in comparing groups leads to different conclusions when using summative scores from rating scales compared with using their corresponding Rasch-based measures. METHODS: A Monte Carlo simulation study was designed to examine between-group differences in the change scores derived from summative scores from rating scales, and those derived from their corresponding Rasch-based measures, using 1-way analysis of variance. The degree of inconsistency between the 2 scoring approaches (i.e. summative and Rasch-based) was examined, using varying sample sizes, scale difficulties and person ability conditions. RESULTS: This simulation study revealed scaling artefacts that could arise from using summative scores rather than Rasch-based measures for determining the changes between groups. The group differences in the change scores were statistically significant for summative scores under all test conditions and sample size scenarios. However, none of the group differences in the change scores were significant when using the corresponding Rasch-based measures. CONCLUSION: This study raises questions about the validity of the inference on group differences of summative score changes in parametric analyses. Moreover, it provides a rationale for the use of Rasch-based measures, which can allow valid parametric analyses of rating scale data.
OBJECTIVE: To investigate whether using a parametric statistic in comparing groups leads to different conclusions when using summative scores from rating scales compared with using their corresponding Rasch-based measures. METHODS: A Monte Carlo simulation study was designed to examine between-group differences in the change scores derived from summative scores from rating scales, and those derived from their corresponding Rasch-based measures, using 1-way analysis of variance. The degree of inconsistency between the 2 scoring approaches (i.e. summative and Rasch-based) was examined, using varying sample sizes, scale difficulties and person ability conditions. RESULTS: This simulation study revealed scaling artefacts that could arise from using summative scores rather than Rasch-based measures for determining the changes between groups. The group differences in the change scores were statistically significant for summative scores under all test conditions and sample size scenarios. However, none of the group differences in the change scores were significant when using the corresponding Rasch-based measures. CONCLUSION: This study raises questions about the validity of the inference on group differences of summative score changes in parametric analyses. Moreover, it provides a rationale for the use of Rasch-based measures, which can allow valid parametric analyses of rating scale data.