作者: Geoff J. McLachlan , David Peel
关键词: Resampling 、 Multivariate statistics 、 Algorithm 、 Carry (arithmetic) 、 Mixture model 、 Mathematics 、 Expectation–maximization algorithm
摘要: We consider the fitting of normal or t-component mixture models to multivariate data, using maximum likelikhood via EM algorithm. This approach requires initial specification an estimate vector unknown parameters, equivalently classification data with respect components model under fit. describe algorithm called EMMIX that automatically undertakes this fitting: including provision suitable values if not supplied by user. The has several options, option carry out a resampling-based test for number in model.