catSurv: Computerized Adaptive Testing for Survey Research

Surveys are expensive — why ask everyone the same questions? catSurv implements computerized adaptive testing (CAT) methods for survey research, dynamically selecting the most informative questions for each respondent based on their previous answers. The package supports a range of item response theory models (latent trait, three-parameter, graded response, generalized partial credit) along with multiple ability estimation and item selection routines. A C++ back-end ensures computational efficiency for large-scale surveys. The methodology is described in Montgomery and Rossiter (2020), Journal of Survey Statistics and Methodology.