スキルテストへの項目応答理論の適用（2）

Abstract

This study investigated the applicability of item response theory to motor skill tests. In particular, we focused on the evaluation of children's motor skills using three Rasch models. Motor skill tests included the number of successes in a given trial, the number of successes in a given time, and a subjective rating regarding the form of the motor skill. We used Andrich's Binomial Trials Model, Rasch's Poisson Counts Model, and Masters' Partial Credit Model, respectively, to derive a synthesized motor skill ability using the item response theory from two or more motor skill test scores. A synthesized motor skill ability was computed by combining these three models and applying them to children's motor skill test items. The sample consisted of 205 kindergarten-aged children, and 8 motor skill items were tested, including (1) side jump, (2) jumping over a 30-cm high rope with hands on a board, (3) kicking and aiming a ball, (4) catching a ball, (5) rolling and aiming a ball, (6) forward roll, (7) side roll, and (8) jumping and turning 90 degrees until a complete 360-degree turn is completed. The following results were obtained:1) The correlational structure among motor skill test items for this sample had high homogeneity and was one-dimensional. High correlations were found between two sets of Difficulty Parameters, which were computed from two randomly divided subgroups. Obtained Difficulty Parameters were not affected by the sample used. 2) Significant correlations were found between synthesized ability parameters and age, standing height, and body weight, indicating that the motor skill ability parameters measured a latent motor skill, which developed with advancing age.3) A high correlation was found between synthesized motor skill ability and the Principal Component Scores computed by Principal Component Analysis. However, ability parameters showed lower correlations between motor skill test items and age and physique than Principal Component Scores. In contrast, high correlations between motor skill ability parameters and motor skill test items were found.Finally, the likelihoods for obtaining a synthesized motor skill parameter and an expected score for each test item were obtained, and an explanation for using these two values was provided.