Trial Statistics: Reading Numbers, Knowing Terms
Valid and reliable farmer-generated information is a cornerstone of Practical Farmers of Iowa. Consequently, PFI has worked to develop practical methods that safeguard the accuracy and credibility of that information. PFI cooperators use methods that allow statistical analysis of their on-farm trials. Chief among these methods are: 1) replication, and 2) randomization. (See figure, a typical PFI trial.) The farming practices compared in a trial are repeated, or replicated, at least six times across the field. Thus trial results do not depend on a single comparison only, but on six or more. The order of the practices, or treatments, in each pair is chosen with a flip of the coin. This randomization helps to avoid unintentional bias.
PFI on-farm trials have been recognized for their statistical reliability. So, while PFI cooperators dont have all the answers, they do have a tool for working toward those answers. With a little coaching, its not hard to do your own replicated research.
When you see the outcome of a PFI trial, you also see a statistical indication of the strength of the difference observed. The following information should help you to understand the reports of the trials contained in this report. The symbol * shows that there was a statistically significant difference between treatments; that is, one that likely did not occur just by chance. We require ourselves to be 95% sure before we declare a significant difference. If, instead of a *, there is a N.S., you know the difference was not significant at the 95 percent confidence level.
A/B Comparisons Many on-farm trials are of a straightforward A versus B type. These trials, which are easy to design and analyze, correspond to the typical experimental question Is alternative B better than, worse than, or the same as my customary practice A? This approach can be used to evaluate individual practices or whole systems of practices.
There is a handy yardstick called the LSD, or least significant difference, that can be used in a trial with only two practices or treatments. If the difference between the two treatments is greater than the LSD, then the difference is significant. You will see in the tables that when the difference between two practices is, for example, 5 bushels (or minus 5 bushels, depending on the arithmetic), and the LSD is only, say, 3 bushels, then there is a * indicating a significant difference.
Multiple Treatment Trials The LSD doesnt work well in trials with more than two treatments. In those cases, letters are added to show whether treatments are statistically different from each other. (We usually use a statistical test called a multiple range grouping.) The highest yield or weed count in a trial will have a letter a beside it. A number with a b next to it is significantly different from one with an a, but neither is statistically different from a result bearing an ab. A third treatment might produce a number with a c (or it might not), and so on.
Economics Average statewide prices for inputs were assumed in calculating the economics of trials. Average fixed and variable costs and time requirements were also used. These can vary greatly from farm to farm, of course. For 1998, the calculations used prices of $2.00 per bushel for corn, $5.20 for soybeans, and $1.10 for oats. Labor was charged at $9.00 per hour.
Some tables show both a treatment cost (which includes relevant costs, but not the total cost of production) and treatment benefit. The treatment benefit is the relative advantage of a practice compared to a check or control treatment or customary practice, which is often assigned a treatment benefit of $0.
If there are no significant yield differences, treatment benefit is calculated solely from input costs. If the yield of a practice is significantly different from that of the control treatment, then that difference in bushels is also taken into account to calculate treatment benefit for the more profitable practice.
A treatment benefit that is a negative number indicates a relative loss. The highest-yielding practice doesnt always have the greatest treatment benefit. You will see that sometimes the additional input costs of a practice outweighs its greater gross return. And in some trials, the least profitable practice is not the lowest yielding.
Here is one more thing to be aware of. Fertilizer shown with dashes between the numbers (18460) means percent by weight of nitrogen, phosphate, and potash in the product. Fertilizer shown with plus signs (18+46+0) indicates pounds per acre of those nutrients in an application.
The results that appear here imply neither endorsement nor condemnation of any particular product or practice. Producers are encouraged to carry out their own trials to find what works in their operations. In reports of trials that involve proprietary products, brand names are included for the purpose of information.