This paper presents an information efficient technique to determine the functional forms of density and regression functions. By maximizing entropy subject to given side conditions, a flexible functional form approach is established. As special cases of this approach, exponential polynomial, Cobb-Douglas, translog, generalized Cobb-Douglas, generalized Leontief, Fourier flexible form, and minflex-Laurent series, etc. can be derived. In a Monte Carlo experiment, the maximum entropy regression method shows good performance in terms of goodness of fit. Boxplots and isoquants produced by the maximum entropy method are compared with those produced by other well-known functions. © 1993.
Ryu, H. K. (1993). Maximum entropy estimation of density and regression functions. Journal of Econometrics, 56(3), 397–440. https://doi.org/10.1016/0304-4076(93)90128-R