This paper illustrates the use of a statistical technique, finite mixture models, to fit the parameters in cumulative prospect theory. For a given decision, some individuals may adopt a gain frame, while others may adopt a loss frame. By using finite mixture models, the best fitting parameters can be obtained for the two subgroups, even though the information about subjective frames was not available. Our application uses two health insurance survey datasets collected by Rand and Chinese State Natural Science Foundation, respectively. The results are compared with previous studies on framing effects and parameterizations of prospect theory.