Expected Utility or Prospect Theory Maximizers? Results from a Structural Model based on Field-experiment Data