Partitioning general probability measures (Q579730)

Consider n probability measures \(\mu_ 1,...,\mu_ n\) on the same measurable space. It is the aim to split the whole sample set \(\Omega\) into n measurable parts \(A_ 1,...,A_ n\) such that \(A_ j\) has nearly the same probability for each distribution. Here the author considers the case when the atoms have not too large mass. Bounds for the accuracy of the partitions are derived. In the classical situation it is known that for nonatomic distributions there exists a partition with \(\mu_ j(A_ i)=1/n\) for each i and j. The problem and the solution is connected with famous names as Fisher, Neyman, Steinhaus, Banach, Dvoretzky, Wald, Wolfowitz, Dubins and others.