What Does Test Plot Replication and Randomization Mean?
Agricultural input sellers often point to test plot trials as a means of selling their product. But how do you know if these trials are accurate?
One way is to ask if the test plots are replicated and randomized. If a trial has both these attributes, a high degree of accuracy with minimal error is ensured.
Replication means repeating a treatment multiple times. In university small-plot trials, a treatment may be replicated three or more times. The more replications a trial contains, the more confident researchers are regarding the results.
Results for each test plot will vary. However, the more a treatment is replicated, the more confident testers can be that representative results occur.
It’s an important way to further boost test plot accuracy. Randomization mixes up patterns between trials. For example, a trial may have treatments replicated may occur on the left side of a field or plot, where the next treatment may reverse the order.
“Soil type, compaction, wind exposure, disease, and shade levels can vary between plots or field,” says Paul Vincelli, University of Kentucky Extension plant pathologist. We take away the possibility of biases through randomization.”
So why don’t all trials replicate and randomize? They’re a pain in the neck to do, particularly on a field-size basis. Repeating treatments and mixing them up takes time and effort to do. You also have to harvest them in the same pattern.
Vincelli notes farmers can still learn from unreplicated and unrandomized on-farm trials. However, they will have a higher probability of misleading results than do replicated and randomized trials.
Some companies will compile on-farm trials in the hundreds or thousands across a wide geographic area. This can reduce variance and narrow error rate by achieving the same effect as replication.
“With these trials, I don’t worry so much about lack of replication, because they have so many numbers,” says Vincelli.
Randomization, though, is another question. Mixing up planting and harvest patterns through plot compilation is difficult to do. Plots that are both replicated and randomized still give the most accurate read on product performance, he adds.