Abstract: The paper approached the difference of the number of selected superior family between the methods of performance level analysis and simple ttest, under the condition of missing plots, missing trees and variance heterogeneity. The accuracy of simple ttest was analyzed. It was put forward that family selection can be carried out by combining performance level analysis and simple ttest. Eleven superior families were selected from the 32 tested families and the real gain for volume was estimated to be 35.8%.