A mean-squared-error-based forward selection methodology is proposed for simultaneous parameter ranking and selection based on the critical ratio r(CCW) [Eghtesadi, Z.; Wu, S.; McAuley, K. B. hid. Eng. Chem. Res, 2013, 52, 12297]. This new technique employs information in the available data set and the operating region of interest to determine the best model with the lowest mean square prediction error. This technique involves relatively simple computations and avoids overfitting of noisy data. This new approach is valuable when data available for parameter estimation arise from correlated experimental designs that make accurate estimation of all the parameters difficult. It is particularly beneficial when the predictions are desired in an operating region that is different from where the data are already available. Monte Carlo simulations of a linear regression example and a nonlinear case study are used to illustrate the effectiveness of the proposed method, demonstrating that results from simulated data agree with theoretical results.