08.09.2022 Unternehmen

IACAT-Kongress zum Adaptiven Testen im September 2022

Im September kommt die globale Fachcommunity zum Thema „Adaptives Testen“ in Frankfurt zur IACAT-Konferenz zusammen. Die dgp-Testentwicklerin Dr. Anna-Lena Jobmann wird dort in der Session „CAT Applications: Ability Testing“ mitdiskutieren und die neuesten Entwicklungen der dgp präsentieren. In ihrem Beitrag geht es konkret um die Konstruktion von adaptivem Testen mit figuralen Matrizen in der Personalauswahl.

Das finale Programm des Kongresses ist seit heute online und kann hier abgerufen werden: https://www.iacat2022.com/program-2/


++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++


Lesen Sie hier das zur Konferenz eingereichte Abstract von Dr. Anna-Lena Jobmann:

Development of a computer-adaptive test of figural matrices (CAT-FM) for personnel
selection

The paper presents the development of a computer-adaptive test of figural matrices
(CAT-FM) for personnel selection. Figural matrices were constructed using an automated item
generator (AIG) based on six different rules with three varying element groups. A total of N =
7838 applicants for German for middle, upper and highest grades of public service are used
for calibration of 210 generated figural matrices in a high-stake situation. Items were
presented in 11 blocks of 15 to 30 items each. Sample sizes for calibrations of item blocks
differed due to organizational reasons. Items are analyzed and stepwise reduced with regard
to internal consistency, item-total correlations, fit to a one-dimensional model, item fit (RMSD)
for the 2pl model as well as differential item functioning. The final item bank consists of 190
items with appropriate 2pl model fit. Evidence of convergent validity is provided by high latent
correlations to numerical and verbal reasoning. Simulations are used to compare different
stop-criteria for practical use of CAT-FM: With approximately 9 items a reliability of Rel = .874
and with 22 items a reliability of Rel = .938 was reached. The item bank allows to measure
reasoning ability with low standard error (SE <= .316) especially for ability levels between -1
and 1. Based on the results of simulations recommendations for practical use of the CAT-FM
are presented and discussed.

Zurück