Classification of Risk Preferences with Elicited Utility Data: Does Functional Form Matter?