<![CDATA[This research is a simulation study using resampling methods to see the effect of sample size, test length, and person-item targeting on separation reliability in the Rasch Model. Simulation conditions were created with several predetermined factors, namely sample size with five conditions (200, 500, 1000, 2000, and 4000 person), test length with three conditions (20, 40, and 60 items), and person-item targeting with five conditions (-2, -1, 0, 1, and 2 logit). The total number of conditions is 75 conditions where each condition is replicated 50 times so that a total of 3.750 data are generated. The data is generated using WinGen software. The results of the separation reliability analysis were analyzed using Winsteps software. The separation reliability criteria set are for Person Separation Reliability (PSR) > 0.80 and for Item Separation Reliability (ISR) > 0.90. The results showed that 75 conditions (100%) resulted in ISR estimates that met the criteria (> 0.90). For PSR estimation, 37 conditions (49%) resulted in PSR estimates that met the criteria (> 0.80) and 38 conditions (51%) resulted in PSR estimates that did not meet the criteria (< 0.80). In addition, PSR estimation is influenced by test length and person-item targeting.]]>
Copyrights © 2024
