Title: A Quadrature Kalman Filter for Estimating Longitudinal Item Response
Models.

Author: Peter van Rijn

Affiliation: Educational Testing Service Global

Abstract: A general method for estimating longitudinal item response theory
(IRT) models is demonstrated.  The method is based on a discrete-time
Kalman filter that uses Gauss-Hermite quadrature to deal with nonlinearity
and non-normality.  The use of quadrature is highly similar to how it is
used in marginal maximum likelihood estimation of standard IRT models,
thereby providing a natural extension of the latter method to longitudinal
settings.  The method originates from engineering, but is seen in other
fields as well, e.g. in health research.  Several illustrations will be
provided starting with the pure time series case.  Possible applications of
the method in multidimensional IRT and adaptive testing are discussed.