EMPIRICAL VALIDATION OF A MARKOVIAN LEARNING-MODEL FOR KNOWLEDGE STRUCTURES

Falmagne (1994) introduced three Markovian learning models for knowled ge structures. Within these models the probability distribution over t he knowledge states of a knowledge structure at one point in time (or the joint distribution for t points in time) is explained as the resul t of a nonobservable learning process on the knowledge structure. In t his paper a first test of the applicability of such a model is describ ed. Estimations for stochastically simulated data show the explanatory power of the model. Reasonable fits to the data and to the original p arameter values are found. The fit deteriorates drastically if the kno wledge structure differs from the knowledge structure for which the da ta were simulated. An application of the model to two empirical data s ets shows that the predicted data deviate significantly from the empir ical data. This suggests that modifications of the model are necessary . (C) 1997 Academic Press.