Event Title

The accuracy of item and ability estimate of concurrent calibration under the Interlaced Anchor Test design administration

Start Date

24-8-2022 7:00 PM

End Date

24-8-2022 8:00 PM

Subjects

Test construction, Testing, Cheating, Evaluation methods

Abstract

The Indonesia’s teacher’s qualification assessment was administered to more than 600,000 participants in the first of three batches consisting of several days of administration. To amp-up the test security and avoid any threats of cheating and test-stealing that could affect the validity of the results, equivalence forms were developed and an interlaced anchored design with a systematic block of missing responses was implemented. To investigate the effectiveness of the design, a simulation was conducted to study the accuracy of the concurrent calibration estimation under two parameters logistic model of the item response theory. Sample sizes were varied with one thousand replications. Root mean square of difference between the true value and the estimates was computed partition into bias and variance. Positive results were obtained and will be discussed in the light of the implication of the design in real conditions.

Geographic Subject

Indonesia

This document is currently not available here.

Share

COinS
 
Aug 24th, 7:00 PM Aug 24th, 8:00 PM

The accuracy of item and ability estimate of concurrent calibration under the Interlaced Anchor Test design administration

The Indonesia’s teacher’s qualification assessment was administered to more than 600,000 participants in the first of three batches consisting of several days of administration. To amp-up the test security and avoid any threats of cheating and test-stealing that could affect the validity of the results, equivalence forms were developed and an interlaced anchored design with a systematic block of missing responses was implemented. To investigate the effectiveness of the design, a simulation was conducted to study the accuracy of the concurrent calibration estimation under two parameters logistic model of the item response theory. Sample sizes were varied with one thousand replications. Root mean square of difference between the true value and the estimates was computed partition into bias and variance. Positive results were obtained and will be discussed in the light of the implication of the design in real conditions.