Looking-backward probabilities for Gibbs-type exchangeable random partitions (Q2345113): Difference between revisions

The authors are focused on the class of Gibbs-type exchangeable random partitions. These are random partitions which arise by sampling from a random probability measure of Gibbs-type, denoted \(\widetilde P_G\) (see [\textit{J. Pitman}, ``Poisson-Kingman partitions'', in: D. R. Goldstein (ed.), Statistics and science: a Festschrift for Terry Speed. Beachwood, OH: Institute of Mathematical Statistics. 1--34 (2003)]). A recent trend is the study of their conditional properties, that is, the study of the estimation of the distribution of certain statistics of an additional sample \((X_{n+1},\dots, X_{n+m})\), to the initially given sample \((X_1,\dots, X_n)\) from \(\widetilde P_G\). The authors point out the existence of a plenty of open problems in the conditions analysis of Gibbs-type exchangeable random partitions. The present study advances some new results, as for instance on the looking-backward species sampling problems in the general case of Gibbs-type exchangeable random partitions and in the special case of the celebrated Ewens-Pitman sample model. The paper is well-motivated and didactically written.