Estimating Yield Distributions with a Stochastic Trend and Nonnormal Errors

Abstract

Abstract Randomness in crop yields can be decomposed into two broad modeling focuses: the estimation of the mean or central tendency of the distribution and the dispersion around that central tendency. We propose modeling the central tendency of the distribution with a stochastic trend model and allowing for nonnonnality within the stochastic trend through an inverse hyperbolic sine distribution. Results are consistent with this construction. First, residuals around the stochastic trend model are found to be non normal. Second, the inverse hyperbolic sine modification of the stochastic trend model corrects both skewness and kurtosis of corn yields.