On the basis of researching the hydrologic correlation regression with least square method, this paper not only puts forward the using condition, limitation and problems existing in the traditional hydrologic regression analyzing method, but also gives a new idea that variables of different types should be differentiated in regression. In consideration of the present hyclrometric fact that the values of non-relevant variables can be obtained infallibly, the regression method based on the weight of observational biases was explored, the double-weighted regression model using least square method was proposed, and the approaches and formulas of bias estimation both for the model and for parameters were presented. The research on estimating the bias of the model and parameters redounds to the improvement in the accuracy and precision for hydrologic variables regression and bias estimation.