On the proximity of distributions in terms of coinciding fractional moments. (Q1412465)

The purpose of the paper is to investigate the following question: If there are assigned fixed values of \(M\) fractional moments of a probability distribution on \([0,+\infty)\), how close are all the probability measures sharing the same \(M\) moments? This question can be formalized also as the problem of estimating the diameter of the class of all probability measures that have the same fixed values of the first \(M\) moments. The closeness of probability measures sharing the same fractional moments can be formulated by the use of maximum entropy distributions, constrained by fractional moments, in the distribution recovering and the consequent convergence in entropy. The author proves such an existence theorem and the convergence in entropy that allows obtaining a uniform bound for any couple of distributions having the same fractional moments.