Mixture of polynomials probability distributions for grouped sample data
This paper describes techniques for developing a mixture of polynomials (MOP) probability distribution from a frequency distribution (also termed grouped data) summarized from a large dataset. To accomplish this task, a temporary dataset is produced from the grouped data and the parameters for the MOP function are estimated using a Bspline interpolation technique. Guidance is provided regarding the composition of the temporary dataset, and the selection of split points and order of the MOP approximation. Good results are obtained when using grouped data as compared to the underlying dataset, and this can be a major advantage when using a decision support system to obtain information for estimating probability density functions for random variables of interest.
B-spline interpolation, Bayesian information criterion, Frequency distribution, Grouped data, Mixture of polynomials
Cobb, Barry R. "Mixture of Polynomials Probability Distributions for Grouped Sample Data." In European Workshop on Probabilistic Graphical Models, pp. 129-144. Springer, Cham, 2014.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)