Losing control: What happens when the control treatment is excluded in nitrogen fertilizer rate trials?

Providing accurate and precise nitrogen (N) fertilizer recommendations remains a significant challenge. To arrive at a recommendation, researchers traditionally conduct hundreds or thousands of field experiments measuring the grain yield response to varying N fertilizer rates. A statistical model is then fit to the data to calculate the agronomic optimum nitrogen rate (AONR)—the fertilizer N rate that maximizes crop yield. We evaluated the impact of excluding individual fertilizer rates on the AONR estimate and its precision using a mixed‐effects quadratic‐plateau model on a previously published dataset of 49 maize ( Zea mays L.) fertility trials with eight N fertilizer rates. When excluding the control treatment (0 kg N ha −1 ), the AONR deviated from the AONR calculated using all eight rates from 0 to 59 kg N ha −1 for individual sites—a larger change than when any other N rate was excluded. Excluding the control treatment also caused the greatest loss in the AONR estimate precision, with an average standard error increase of +43% without the control compared to +23% for other N rates. Furthermore, our simulations confirmed these findings and showed the largest losses of accuracy and precision when the control treatment was excluded. These trends in bias and precision persisted with different simulated experimental designs. Our results demonstrate the importance of including the control treatment in N fertility trials designed to estimate the AONR. Therefore, we recommend including a control treatment for a more precise N fertilizer recommendation program, especially in on‐farm research.