A Classification Problem in which Information about Alternative Distributions is Based on Samples