Reliable test plots provide answers

We recommend that you should have a test plot for varieties you intend to use and always include newer varieties in the plots. You don’t have time to test every treatment you use, but seed genetics is one of the most important inputs in your control.

Key Points

Every farmer should plant a test plot of varieties they intend to use.

If you can only plant one, choose a variety plot.

Do what you must to stay unbiased.

Single strips are OK for observations, but replication can measure variability and provide more dependable data. Replications must be randomized. Each variety gets an equal chance of being on a certain piece of ground next to a certain variety as any other. Randomization helps remove personal biases and find real differences.

Stay unbiased

Assume there is “no difference” among varieties, and let the data guide you toward the truth, rather than trying to change the data to get what you want. Some say one location is not enough. I say it’s much better than not planting any test. Combine it with unbiased sources, such as university trials.

Always look for the LSD, or least significant difference, in university tables. It measures variability, which may be caused by real differences in varieties, soil types or experimental error. More uniform is more reliable.

It’s not a big deal if a hybrid tops one plot. Look for hybrids that are among the top tier in several locations.

We don’t always get the results we expect, but we publish them anyway, without bias.

Nanda is an agronomic crops consultant and director of genetics and technology at Seed Consultants Inc. Contact him at 317-910-9876, or

Simple tips for on-farm test plots

‘The truth — did a treatment cause a response or not — always exists. It’s just our job to find it,” says Emerson Nafziger, Extension agronomist at the University of Illinois. “We aren’t in the business of doing nice trials; rather,
applied research is the business of trying to say something when we’re done.

“With ‘yes-no’ type inputs — for example, to use a fungicide or not — assign a treatment randomly to one strip of paired strips,” he says. This should be done early to prevent bias.

Keep on-farm research simple. Make the strip wide enough to allow for borders. Randomize within each repetition, use four to eight pairs of repetitions per location, and keep accurate records.

A “significant” effect means the treatment likely caused it, Nafziger notes. Even so, it doesn’t mean the treatment will pay.

“Non-significant results can be obtained from no effect of the treatment, or from so much variability that we can’t separate a treatment effect from background noise,” he adds.

The agronomist also advises against cherry-picking data. “If you need a certain answer, why bother to do the work?” he asks. “Our point is not to find significance, but to figure out what happened and where we go from there.

“There really aren’t any shortcuts,” he concludes. “Statistics do not substitute for the large amount of data and keen observation that good on-farm research always requires.”

Jennifer Shike, University of Illinois news writer, contributed to this story.


Emerson Nafziger

This article published in the March, 2011 edition of INDIANA PRAIRIE FARMER.

All rights reserved. Copyright Farm Progress Cos. 2011.