A Functional Approach to Modeling Test Data
The central problem in psychometric data analysis using item response theory is to model the response curve linking a level θ of ability and the probability of choosing a specific option on a particular item. Most approaches to this problem have assumed that the curve to be estimated is within a restricted class of functions defined by a specific mathematical model. The Rasch model or the three-parameter logistic model for binary data are best known examples. In this chapter, however, the aim is to estimate the response curve directly, thereby escaping the restrictions imposed by what can be achieved with a particular parametric family of curves. It will also be assumed that the responses to an item are polytomous, and can involve any number of options.
KeywordsItem Response Theory Latent Trait Regression Spline Kernel Smoothing Multiple Choice Exam
Unable to display preview. Download preview PDF.
- Abrahamowicz, M. and Ramsay, J.O. (1991). Multicategorical spline model for item response theory. Psychometrika 56, 5–27.Google Scholar
- Eubank, R.L. (1988). Spline Smoothing and Nonparametric Regression. New York: Marcel Dekker, Inc.Google Scholar
- Hastie, T. and Tibshirani, R. (1990). Generalized Linear Models. London: Chapman and Hall.Google Scholar
- Ramsay, J.O. (1993). TestGraf: A Program for the Grapical Analysis of Multiple Choice Test Data. Unpublished manuscript. McGill University, Montreal, Canada.Google Scholar
- Ramsay, J.O. and Winsberg, S. (1991). Maximum marginal likelihood estimation for semiparametric item analysis. Psychometrika 56, 365–379.Google Scholar
- Wang, X. (1993). The analysis of test data using smoothing splines. Unpublished thesis, McGill University, Montreal, Canada.Google Scholar