Correlation Dimension as a Proxy for the Number of Parameters Required to Model Risk of Groundwater Arsenic Contamination: An Assessment Using Ordinary Logistic Regression