Evaluation of eight forest fertilizer response estimators by means of a simulation study

Eight estimators of total fertilizer response (simple difference, ratio, covariance, and regression estimators) commonly used to determine the results of fertilization experiments were evaluated using a simulation study. An individual tree model based on equations developed for Douglas-fir (Pseudotsugamenziesii (Mirb.) Franco) was used to project the growth of fertilized and unfertilized plots under several different experimental conditions. The effects of number of plots, plot size, initial plot variability, stand age at the time of fertilization, length of response period, and differing growth responses to fertilization on the relative performance of the estimators were evaluated. Covariance and regression estimators provided the best estimates as long as there was sufficient data to determine model parameters. In almost all cases measures of growth were preferable to measures of yield. For the simple difference and ratio estimators strong linear relationships were observed between bias and pre-treatmen...