A Comparison of Non-Linear Regression and Weighted Least Squares for Predicting Visibility in Germany.

Abstract : Two methods of parameter estimation for the two parameter Weibull distribution are investigated using data on visibility from Germany. They are: Non-Linear Regression and Weighted Least Squares. Non-Linear Regression gave superior results for both constants and variables models when measured either by overall RMS or by sample reuse. However, since the Weighted Least Squares gave similar results and utilized only about one-tenth the computer time, the authors recommend that it might be used to investigate other models of visibility for other countries. (Author)