An efficient regression-based outlier detection method has been described in this paper. Estimating a correct regression model is an important part of estimation-based outlier detection. In the conventional method, the parameter estimation of the regression model was performed using measurements among neighbor sensors. This research proposes a weighted summation approach for making a synthesized independent variable from the observed values. A reliable regression model can be estimated from the synthesized signal. Finally, accurate outlier detection can be performed using the residual analysis of the regression model. The proposed method was successfully applied in real sensor measurements in a dam.