Response Time models (39)
Chair: Margo Jansen, Wednesday 22nd July, 11.45 - 13.05, Lowercroft, School of Pythagoras.
Jochen Ranger and Christof Schuster, Psychologische Methodenlehre, Justus-Liebig-Universitat Giessen, Germany. A model for the joint distribution of dichotomous item responses and discrete response times. (026)
Margot Jansen, University of Groningen, The Netherlands. Modeling response behavior on tests with time restrictions. (146)
Maretha Prinsloo, Magellan Consulting (pty) Ltd & Cognadev, Northcliff, South Africa. Challenges in cognitive assessment. (003)
ABSTRACTS
A model for the joint distribution of dichotomous item responses and discrete response times. (026)
Jochen Ranger and Christof Schuster
This model for considering item responses and response times jointly is based on two latent factors: ability and speed. Conditional on these factors, responses and response times are considered independent. Whereas item responses are assumed to depend on ability only, the response times depend on both the ability and the speed-factor. Based on the proportional hazard model, the probability of responding before time t is related to the factors via the complementary log-log link. By using discrete response times, assumptions about the underlying distribution can be avoided. Feasibility of parameter estimation by marginal maximum likelihood is demonstrated by a small-scale simulation study. The effect of considering response times in addition to item responses on the precision of ability estimates is discussed.
Modeling response behavior on tests with time restrictions. (146)
Margot Jansen
Many mental tests are given under more or less severe time restrictions. Subjects which are taking a test are not only differing in ability but also in mental speed, and some might run out of time towards the end of the test. When both response times and item responses are collected, it is possible to measure accuracy and mental speed using an IRT model combining both aspects, when individual (item) response times are not available we might use other item or subject characteristics to represent the speed factor, such as item order, the number of items not reached. We consider a simple IRT model where we introduce a second person parameter, which can be viewed as a negative "gain" parameter, which models a decreasing probability of solving an item after answering each successive item. Examples will be given.
is Challenges in cognitive assessment. (003)
Maretha Prinsloo
Cognitive assessment informs decisions and practices in both educational and work environments. Guidelines are thus required for purposes of selection, placement, career guidance, thinking skills development, succession planning, group compilation, job structuring, etc. IQ tests and Structured interviews, amongst others, have traditionally been used to assess intellectual functioning. IQ tests focus on the measurement of linear-causal, logical-analytical, convergent thinking in structured and content specific environments. Previous exposure plays a key role. IQ tests do not predict dynamic constructs such as learning potential and are culturally loaded. Results based on Structured interviews are affected by interviewer skill, rapport, hindsight, verbal eloquence, personality factors and potential halo effects. In this paper a new paradigm for the measurement of cognition is presented and the methodology on which two of the related tools, namely the Cognitive Process Profile (CPP) and the Learning Orientation Index (LOI) are based, is discussed. Both techniques are computerized simulation exercises, which externalizes and tracks micro thinking processes according to approximately 10,000 feedback loops. The results are interactively integrated to provide a comprehensive analysis of the person’s Cognitive Styles, “Brain profile” preferences, preferred unit of information, preferred pace of problem solving, most suitable work environment, strengths and development areas, developmental guidelines and Learning potential. Both the LOI and CPP have been developed to accommodate for cross cultural differences.