The application of multivariate data analysis, a method for coping with multi-colinearity among independent variables in analyzing coastal water quality data, is presented. This study investigates the statistical regr...The application of multivariate data analysis, a method for coping with multi-colinearity among independent variables in analyzing coastal water quality data, is presented. This study investigates the statistical regression modeling of FIB （fecal indicator bacteria） concentrations at the outlet of Talbert Marsh in Orange County, California. The multivariate data modeling utilized FIB and physical variables measurements （n = 5,580） collected during a series of longitudinal study of the Talbert Marsh. For the statistical prediction modeling in predicting the FIB concentrations at the outlet of the Talbert Marsh, multivariate analysis techniques such as PCR （principal components regression）, PLS （partial least-squares） regression and SVM （support vector machine） regression were adopted. Statistical modeling results suggest that the statistical modeling predictions are all fell within the reasonable range of actual measurement data. In addition, it is indicated that the accuracy of SVM regression for predicting FIB concentrations at the Talbert Marsh outlet is better than that of other models.展开更多
文摘The application of multivariate data analysis, a method for coping with multi-colinearity among independent variables in analyzing coastal water quality data, is presented. This study investigates the statistical regression modeling of FIB （fecal indicator bacteria） concentrations at the outlet of Talbert Marsh in Orange County, California. The multivariate data modeling utilized FIB and physical variables measurements （n = 5,580） collected during a series of longitudinal study of the Talbert Marsh. For the statistical prediction modeling in predicting the FIB concentrations at the outlet of the Talbert Marsh, multivariate analysis techniques such as PCR （principal components regression）, PLS （partial least-squares） regression and SVM （support vector machine） regression were adopted. Statistical modeling results suggest that the statistical modeling predictions are all fell within the reasonable range of actual measurement data. In addition, it is indicated that the accuracy of SVM regression for predicting FIB concentrations at the Talbert Marsh outlet is better than that of other models.
