The paper describes the approach used for the calibration of a price-endogenous programming model, developed for the agricultural sector of the region Khorezm in Uzbekistan. Extensive datasets from farm surveys were used to parameterize the model, which nevertheless tended to over-specialization and failed in general to replicate the observed levels of primal model variables. Calibration of the model with “Positive Mathematical Programming” approaches was not satisfying as the additional cost terms introduced to replicate the observed situation were in many cases not plausible and deviated substantially from any available information on cost structure of the agricultural production activities in the study region. After revising the survey data it became obvious that the variances of technical parameters of the model, namely the input coefficients, were significantly larger than any other used set of information. Consequently, instead of introducing additional cost terms, we decided to estimate the technology parameters, such that the observed situation was replicated and the Kuhn-Tucker conditions for an optimum in this point were fulfilled. The estimation was based on a cross entropy approach. The needed support points and prior distributions of the technology coefficients and dual values were drawn from survey data and additional sources of information, such as expert interviews. The result is a calibrated agricultural sector model with a technology representation that was derived by systematical exploitation of relevant data sources for the study region.