ON RAPID ASSESSMENT METHODS USING STATISTICAL MODELING: MULTIPLE LEAST SQUARES REGRESSION VS. LOGISTIC REGRESSION
Abstract
Objective. To introduce new methodology is assessing recreational water quality use and associated effects on Health Background. There is a need to develop rapid assessment of bacterial water quality. To this end many statistical models have been published mostly using-environmental variables to predict concentrations of a particular FIO. The majority of these statistical models have used Multiple Least squares regression in which the major indicator of the goodness of fit of these models have largely depended on the R2 value, which to date have been quite low. Since Beach management decisions have to be dichotomous in nature (Open/Close Beach) we explored the use of the Multiple logistic model in relation to the Multiple Lease Squares approach. Methods. 668 samples were utilized in this analysis. 10 major environmental variables and several FIO’s were collected on each sample date. Both types of models were run on these data. Results. Our Best Multiple Least Squares Regression was computed with a R Square value of 0.26, while the Multiple Logistic Regression Model yielded a maximum Sensitivity of 72.9% and a maximum Specificity of 65.9% at a cut point = 0.1. A backward selection routine was used in both the Logistic and Least Squares Model. Conclusion. Since the Logistic regression yields a much less nebulous goodness of fit statistic coupled with the fact that the Beach Managers decision is a dichotomous one, more attention should be paid to research using the Multiple Logistic Model. Grants. Data collected during grant: $15000.00 Center of Excellence Oceans and Human Health Center, University of Miami.
ON RAPID ASSESSMENT METHODS USING STATISTICAL MODELING: MULTIPLE LEAST SQUARES REGRESSION VS. LOGISTIC REGRESSION
POSTER PRESENTATIONS
Objective. To introduce new methodology is assessing recreational water quality use and associated effects on Health Background. There is a need to develop rapid assessment of bacterial water quality. To this end many statistical models have been published mostly using-environmental variables to predict concentrations of a particular FIO. The majority of these statistical models have used Multiple Least squares regression in which the major indicator of the goodness of fit of these models have largely depended on the R2 value, which to date have been quite low. Since Beach management decisions have to be dichotomous in nature (Open/Close Beach) we explored the use of the Multiple logistic model in relation to the Multiple Lease Squares approach. Methods. 668 samples were utilized in this analysis. 10 major environmental variables and several FIO’s were collected on each sample date. Both types of models were run on these data. Results. Our Best Multiple Least Squares Regression was computed with a R Square value of 0.26, while the Multiple Logistic Regression Model yielded a maximum Sensitivity of 72.9% and a maximum Specificity of 65.9% at a cut point = 0.1. A backward selection routine was used in both the Logistic and Least Squares Model. Conclusion. Since the Logistic regression yields a much less nebulous goodness of fit statistic coupled with the fact that the Beach Managers decision is a dichotomous one, more attention should be paid to research using the Multiple Logistic Model. Grants. Data collected during grant: $15000.00 Center of Excellence Oceans and Human Health Center, University of Miami.