Based on the sum of regression residual square (Q) and the criteria for choosing some of undependent variables, a choice criterion for improving the forecast accuracy is discussed. Also, the effectiveness of these criteria is examined and compared on the basis of skill scores in the forecast experiments