Comparison of Models for the Linear Regression Analysis of Environmental Monitoring Datasets

The least squares regression minimizing deviations only in dependent variable is not suitable for the regression analysis of environmental monitoring datasets, which are all random variables. Three two-variable linear models, i.e., the least squares regression, the reduced major axis regression, and the least normal square regression, were compared for the regression analysis of anions and cations in rain water samples. The results shown that the reduced major axis regression, rather than the others, was likely to be the model of choice for the regression analysis of random datasets, and a higher value was obtained for the regression coefficient b, showing a better relationship between the variables.