By Jeff Miller, Miller Research LLC
It seems that there is no shortage of products on which crop producers are asked to spend their hard-earned money. Every year potato growers are presented with what seems like an endless array of fertilizer and pesticide choices. And if you are going to spray a pesticide, you can find any number of adjuvants which promise to improve the performance of your application. If that were not enough, different vendors present new ideas for modifying equipment or irrigation water, or anything else related to farming.
How is a person supposed to judge which inputs or upgrades are important and cost-effective, and which are not? Unbiased agricultural research results from an unbiased third party can help you sort through the plethora of information. However, not all research is useful and it is important to know how to interpret research results properly.
This example illustrates the point. What if I told you that using Miller’s Marvelous Masterpiece (MMM) will increase your potato yield, and that I had research to prove it? I conducted a trial where the grower standard practice resulted in a yield of 612 cwt/acre and the MMM produced 626 cwt/acre. The MMM provided an increase of 14 cwt/acre, and based on a contract price of $8/cwt, you just gained an extra $112/acre. Now because I’m a nice guy, I am only going to charge $30 for the MMM program. Subtracting the $30 from $112 gives you a net gain of $82/acre. That $82 is a 2.7 times return on your investment. Have I convinced you to buy this from me yet?
What if the scientist who conducted the trial said that the differences were not “significant”? I have presented a lot of data through the years at different meetings, and on more than one occasion I have heard people say, “That difference may not be significant to you, but it is significant to me.” The word “significant” from a scientific point of view relates to the confidence you have that the results will be repeatable. Put another way, if the scientist were to go out and do the test over or if the grower was going to do the treatment in a field, how likely would they be to see the same result? Statistical tests can be done to answer that question.
Research Trial Must-Haves
The way that a research test is conducted is critical in order to get results that are of any value. Key elements of a good research trial include:
- Applying the different treatments to a uniform crop population
- Replicating the treatments as many times as is feasible
- Assigning the treatments randomly
- Evaluating the treatments in an unbiased manner
All elements are important in order to understand if a treatment is truly having an effect.
The purpose of working with a uniform crop population should make sense. I had an experience once where a grower applied product A on the north side of the pivot and product B on the south. At the end of the year, the level of disease was the same on both sides. The grower’s thought was that product B was the “winner” because the south side of the field typically had more disease. In the end, nothing was learned from this comparison.
Replicating or repeating treatments is also critical, and this is why field splits are not always useful. If a treatment is consistent at causing a specific result (such as increasing yield), then that should happen whether or not I apply it on the north, south, east or west side of the field. In agricultural research, treatments are typically repeated in four to six separate blocks (replications). In an effective trial, the different blocks will be uniform.
Placing the treatments within each block randomly helps to avoid unforeseen problems that may arise due to position in the field. For example, suppose a researcher had four blocks where treatments were being tested in the field. The blocks started at the mainline and stretched north through the field. This researcher always put treatment A at the beginning of each block near the mainline. At the end of the trial, the yield in treatment A was the lowest. The researcher would then be tempted to conclude that treatment A was not effective. But what if drainage from the irrigation lines adversely affected crop growth near the mainline? The poor growth in treatment A was not the result of the treatment, but rather the result of an external factor. This problem would have been accounted for had the treatments been placed randomly.
Finally, treatments need to be evaluated in an unbiased manner. I have evaluated several field demonstration plots where somebody with a vested interest in the results claims to see subtle differences. Often people see what they want to see, and it is common for personal bias to interfere with objective analysis. To overcome this, evaluations should be made without knowing the treatment which is being evaluated. This is usually done by assigning a code or number to each plot. When the researcher goes to the field, results are recorded for each coded plot. After the results are analyzed, the plots are then decoded in order to objectively analyze the data. If the number of treatments is small, it may not be possible to hide the treatment identities completely. In this case, you could ask somebody else who does not know what was done to make the evaluation.
Looking at the Data
So let’s go back to the MMM product I am trying to sell you. I utilized the elements of a good agricultural research trial and I got the results shown in Table 1. Even though the average MMM yield increase was 14 cwt/acre, that increase was not observed in each block. In block 2 the MMM treatment was not better than the grower practice, and in block 3 both were very similar.
|Grower Standard Practice||Miller’s Marvelous Masterpiece|
Table 1. Comparison potato yield in cwt/acre of the grower standard practice to Miller’s Marvelous Masterpiece.
We don’t have the time to go into all the details about the statistical evaluation of the data, but suffice it to say the end result was that there was a 55 percent likelihood that MMM would increase potato yield. That is not a very good likelihood that MMM will work for you. As much as I would like you to buy it from me, it probably won’t work as well as I hope.
In summary, unbiased agricultural research data can be used to help you decide if you should spend the money on products or practices. If the data do not support the use of a product, you are better off not using it or doing a trial of your own following the elements of a good research design. If you are not sure how to do that, contact your local extension specialist to help you design a test.