In nitrogen fertilization trials of maize (Zea mays L.), estimation of both optimum nitrogen fertilizer and maximum grain yield has been of interest to researchers. This requires assessment of statistical models fit to yield data in response to N fertilizer rates before selecting a single model that best fit data. We assessed linear-plateau (L-P), quadratic-plateau (Q-P), quadratic (Q), exponential, and square root models describing four cultivar grain yield responses to five N fertilizer rates (0, 40, 80, 120, and 160 kg N fa -1). This was based upon data from a two-year field experiment conducted in 2001 and 2002. Averaged over cultivars, coefficients of determination, R2 were quite close and considerably high (0.873 to 0.887) for all models, except for the exponential model that yielded fairly low value (0.582). The latter model gave a root mean square error (RMSE) equals 465 whereas the others resulted in values ranged from 248-262. Despite of the closeness of both R2 and RMSE of four models, they resulted in a wide range of predicted optimum N fertilizer rates. The square root model gave absurd values, and the exponential model gave the lowest (58 -76 kg N fa |
