A logistic growth equation with time and location varying parameters was used to model corn response to applied nitrogen. A nonlinear dummy-variable regression model provided a parsimonious representation of site and time effects on parameter values. The model was used to test for the equality of the mean marginal product of nitrogen fertilizer between locations on the coastal plain of North Carolina. Monte Carlo simulation and bootstrap simulation were used to construct finite sample covariance estimates. Results support rejection of the hypothesis that mean marginal products are equal when nitrogen is applied at 168 kg/ac. A comparison of bootstrapped errors and asymptotic errors suggests that results based on asymptotic theory are fairly reliable in this case.