Abstract:Regression is usually used to determine the relation among some testing parameters. Experimental data of rock mechanics are often dispersive,and the reliability of the regression result must be evaluated. In this paper,two methods to review the error of regression are given on the Coulomb criterion. One is to regress again after decreasing or repeatedly using some data. The other is to do‘hypothetical experiments’after arranging the specimens in various confining pressures. A certain range of the rock parameter like internal friction angle with certain reliability can be given. The effect of confining pressure range on the regression results of strength criterion is also discussed.