Making the Most of On-Farm Trials

GPS-enabled equipment has greatly simplified the logistics for conducting on-farm trials. Farmers often rely on local agronomists to help in setting up, monitoring, harvesting, and analyzing results. They form the basis for some great discussions at winter meetings. Plots take time, and they must be set up correctly to produce meaningful results. Replication of treatments is critical to determine if differences between plots are due to treatments or due to natural field variation. The trial design is important whether applied to strip trials or learning block trial design. The graphic here illustrates how plot length and actual yield affect the ability to show ‘real’ differences in treatments (LSD). For example in a 3 replication trial, increasing the length from 250 ft to 1000 ft reduced the LSD from 10 bu/ac to about 5 bu/ac. It also illustrates in a simple 2 treatment comparison, the power of replication. Four to six replications are generally necessary for simple 2-3 treatment comparisons in hopes of showing meaningful results.  Actual yields also make a difference in how long a plot needs to be. Longer plots generally give less trial variability and increase confidence.
Image: Effects of replication number and plot length on the Least Significant Difference (LSD) values at the 5% statistical probability level based on data from 10 field uniformity trials in Idaho, Oregon and Washington with less than 5 bu/acre yield variance. (Adapted from Wuest et al, 1994).

A Trial without replication is a Demonstration