Sample Size and Statistical Conclusions from Tests of Fit to the Rasch Model According to the Rasch Unidimensional Measurement Model (Rumm) Program in Health Outcome Measurement.

Sample size is a major factor in statistical null hypothesis testing, which is the basis for many approaches to testing Rasch model fit. Few sample size recommendations for testing fit to the Rasch model concern the Rasch Unidimensional Measurement Models (RUMM) software, which features chi-square and ANOVA/F-ratio based fit statistics, including Bonferroni and algebraic sample size adjustments. This paper explores the occurrence of Type I errors with RUMM fit statistics, and the effects of algebraic sample size adjustments. Data with simulated Rasch model fitting 25-item dichotomous scales and sample sizes ranging from N = 50 to N = 2500 were analysed with and without algebraically adjusted sample sizes. Results suggest the occurrence of Type I errors with N less then or equal to 500, and that Bonferroni correction as well as downward algebraic sample size adjustment are useful to avoid such errors, whereas upward adjustment of smaller samples falsely signal misfit. Our observations suggest that sample sizes around N = 250 to N = 500 may provide a good balance for the statistical interpretation of the RUMM fit statistics studied here with respect to Type I errors and under the assumption of Rasch model fit within the examined frame of reference (i.e., about 25 item parameters well targeted to the sample).