The problem of metric identification of probability density distribution for a random value is considered. It is shown that the accuracy is determined both by the sample size and by the number of intervals into which the domain of its determining is divided.