This study is to investigate the accuracy of IRT ability parameter estimates based on summed score
EAP method under the polytomous IRT model through several simulation conditions, comparing traditional
EAP method based on item response pattern. The former has an advantage in that it can produce
convincing scale scores to the public by estimating one-to-one ability estimates with the number correct
scores. It also has an disadvantage, however, ignoring the information the item response patterns have.
Even though it has the problem of losing information, this study expects to find the degree of estimation
accuracy is similar to the latter. For this, the study compares the accuracy of recovery of true ability
parameter under the several simulation conditions, and presents the results with the value of MSE, SB, and
VAR. The results show that the latter produces a little bit smaller value of MSE, SB, and VAR than the
former. However, the difference usually happens at the second decimal places. Moreover, such results of
both two EAP have very small difference values close to 0 on the ability parameter scale between ?2
and 2. Therefore, the results imply that the use of EAP based on summed score can be the promising
alternative ability estimation method under the operation of an actual testing program.
Key words : Polytomous item response theory, EAP method based on the summed score, EAP method