How Well Does Monte Carlo Simulation Perform in Generating Datasets Under the IRT 3PL Model?

• In Event: AERA Roundtable Session Sunday 1:45 pm - Gold 3
  In Roundtable Session: RT1: Measurement and Related Issues in Quantitative Research Designs, Analysis, and Simulation

Abstract

This study aimed to test the performance of Monte Carlo simulation in generating datasets under the 3PL IRT model with varying sample sizes, item numbers, and replications. Performance was evaluated using estimation bias and error, correlations between initial and estimated parameters, and the percentage of estimates within defined ranges. Results showed that accurate item parameter recovery is possible when sample size exceeds 1000 (preferably 2000) and item number is 40 or more. Under these conditions, as few as 10 replications may suffice. However, even with sufficient sample size and item number, the c-parameter often diverged from its initially defined value.

Author

• Ergul Demir, Ankara University