In this paper, we present a study on using weighted total least squares method for parameter estimation of errors-in-variables models with quadratic regressors. The statistics of error is analyzed to fill in the gap between basic assumptions in weighted total least squares and our case. A modified Cramér-Rao lower bound is introduced for error quantification in the proposed method. We perform evaluations based on simulations with comparisons to standard least squares and generalized total least squares. Numerical results show that the proposed method outperforms the others in terms of estimation accuracy.