The comparison of descriptive statistical parameter estimation stability using raw scores and rasch model

Abstract

This research aims at analyzing the comparison of descriptive statistical Parameter Estimation stability using raw scores and Rasch model. The empirical data were the responses of the 12th Grade Science Students of Senior High School on Science Literary Test based on the integrated mathematics and natural sciences conducted at SMAN 2 and SMAN 3 Tegal, Central Java. This research employed a bootstrapping method assisted with SPSS version 21, while the Rasch model was assisted with R program version 3.6.3 eRm package Version 1.0.1. The parameter stability estimation was seen from its error standard and bias scores. The scores using the Rasch model was proven giving higher stability when compared to that using the raw scores in its descriptive statistical parameter estimation both from its error standard and bias aspects. Based on the error standard used, it showed that the mean and the standard deviation estimation when using the Rasch model scores was around 8 times more stable when compared to that when using the raw scores, while its median estimation were 16-18 times. Due to the use of bias measurement, it showed that the mean estimation when using the Rasch model scores was 6-10 times more stable when compared to that using the raw scores, while its median estimation and standard deviation were respectively 43-282 times and 7-10 times more stable.