Rasch Model Calibrations with SAS PROC IRT and WINSTEPS.

Area of Science:

• Educational measurement
• Psychometrics
• Statistical software performance

Background:

• WINSTEPS is a standard software for Rasch model calibrations.
• SAS/STAT's PROC IRT procedure now offers Item Response Theory (IRT) analysis, including Rasch models.

Purpose of the Study:

• To compare the performance of SAS PROC IRT against WINSTEPS for calibrating dichotomous and polytomous Rasch models.
• To evaluate PROC IRT as a potential alternative to WINSTEPS for IRT analysis.

Main Methods:

• A simulation study was conducted to assess software performance.
• Key metrics included convergence rates, run times, item parameter estimates, and ability estimates.
• Variations in test length and sample size were examined.

Main Results:

• Performance comparison focused on accuracy and efficiency of parameter estimation.
• Analysis included dichotomous and polytomous Rasch models under varying conditions.

Conclusions:

• The study provides insights into the viability of PROC IRT for Rasch modeling.
• Results inform choices between WINSTEPS and PROC IRT for research and large-scale assessments.