Ability Assessment

The benefits of adaptive testing for a practitioner’s everyday consulting: Either branched or tailored testing with the Adaptive Intelligence Diagnosticum (Keynote)

Klaus D. Kubinger, Universität Wien, Austria

Adaptive testing is psychometrically well-known since more than half a century, but there are hardly psychological tests on practitioners' disposal which are conceptualized accordingly for their everyday case consulting. However, the Adaptive Intelligence Diagnosticum (AID; as an English edition by Kubinger, 2017, as the last German edition by Kubinger & Holocher-Ertl, 2014) actually applies adaptive testing for a widely used intelligence test-battery (for 6 to 16 years old children und juveniles). Admittedly, developmental efforts are high and expensive as at least twice as many items have to be calibrated according to some IRT-model than conventionally administered tests need – calibrated e. g. according to the Rasch model. But efficiency in test administration pays: Applying branched adaptive testing AID has proven already more than thirty years ago to save the administration of half the items when trying for the same measurement accuracy than a conventional test; and applying tailored adaptive testing, there are further three to four items (out of 15 with branched testing) on average which administration can be saved for each subtest of the AID (Kubinger & Spohn, 2017). Obviously, that means saving a lot of practitioners' time, besides saving a good deal of the examinees' energetic-motivational power. The presentation focuses on the psychometric foundation of adaptive testing so on demonstrating AID's attractiveness for practitioners.

Date: 3 May 2021, 3:30 p.m. (Berlin time)

Register: Send email to eapa.online.event@gmail.com and state your name, the event(s) you want to attend and the email address we can contact you with.

Methodological advances in psychological test norming (Symposium)

Lieke Voncken, Tilburg University, The Netherlands

Psychological tests, like intelligence tests, are often used in the diagnosis and selection of individuals. The raw test scores themselves are typically not informative, so a reference population is used for interpretation. For instance, intelligence test scores are typically compared to the scores of the general population of the same age as the testee. In modern psychological test norming, the normed scores are obtained by estimating a regression model in which (part of) the raw test score distribution is related to predictor(s), such as age.
In this symposium, different regression-based norming – also called continuous norming –approaches will be discussed. We will focus on norm estimation via quantile regression, regression via Taylor polynomials, and distributional regression. In quantile regression, specific percentiles are modelled as a linear combination of predictor(s). In regression via Taylor polynomials, the raw test scores are regressed on polynomials of the normed scores and predictor(s). In distributional regression, moments of the raw test score distribution are estimated as a function of predictor(s).
The application of distributional regression will be illustrated for the Student Perception of Affective Relationship with Teacher Scale (SPARTS). In addition, we will discuss how normed scores can be estimated more efficiently by using prior norm information. Regression-based norming might require large normative samples to estimate the normed scores properly, and using prior norm information can alleviate this burden. Finally, we will provide practical implications for test developers and discuss our directions for future research.

Contributions:

• Robust and efficient norming with penalized quantile regression (Marieke Timmerman)
• Making norming robust and simple (Alexandra Lenhard)
• Norming the Student Perception of Affective Relationship with Teacher Scale (SPARTS) (Andries van der Ark)
• More efficient norm estimation by using prior norm information (Lieke Voncken)

Date: 3 May 2021, 4:30 (Berlin time)

Register: Send email to eapa.online.event@gmail.com and state your name, the event(s) you want to attend and the email address we can contact you with.