The paper reviews the empirical problem of estimating state-contingent production functions. The major problem is that states of nature may not be registered and/or that the number of observation per state is low. Monte Carlo simulation is used to generate an artificial, uncertain production environment based on Cobb Douglas production functions with state-contingent parameters. The parameters are subsequently estimated based on different sizes of samples using Generalized Least Squares and Generalized Maximum Entropy and the results are compared. It is concluded that Maximum Entropy may be useful, but that further analysis is needed to evaluate the efficiency of this estimation method compared to traditional methods.