Use good data to evaluate products

Decisions on what variety, nutrient or crop input product to buy are improved with good data. When looking for data, here are a few clues as to the quality of the data set:

Replication. The more sites the better. Having treatments and products tested across Western Canada will give a better idea of how the product performs under different soil and climactic conditions. Making decisions from one site will not give a good idea of performance consistency over a range of conditions and environments.

Multiple site years. Similar to replication, having products tested over multiple years will also give an idea of how the product performs year over year. Conditions will never be the same year over year, and looking at product performance over a wide range of years will give a better idea of product performance over a wide range of conditions, such as too wet, too dry, too hot, etc.

Statistical Information. Look for the following details when analyzing data:
—Coefficient of variation (CV). The larger the CV, the more variability there was in the experiment. Conversely, a lower CV is a reflection of the reliability of the experimental results. Variation comes from two main sources: differences between treatments due to product applied, and lack of uniformity where the trial is conducted. Typically, a CV value of 15 or above would mean there was too much variability in the trial, and the data may not be valuable.
—Least Significant Difference (LSD) is used to indicate when data values represent treatment differences with 95% certainty. Say for example the LSD value in a variety trial is 5.0 bu./ac. If the yield difference between two varieties in the trial is 8.6 bu./ac., we would be 95% certain that the treatments were different. If the yield difference between two varieties is 3.6 bu./ac. (less than the LSD value), we cannot conclude that the treatments are different from each other. It could just be random chance that one had a higher result than the other.

Chance of a response. When comparing a treatment to an untreated check, if there is a response at more than half the sites, there is a probable chance of getting a response by introducing the practice. For variety comparisons, if variety “A” yields higher than variety “B” at more than 50% of the sites, the greater the chance that variety “A” will produce higher yields for farmers. This will often be presented as “% wins”.

Consider the data source. There are many sources of data available to producers. If mentioned, yield data from a weigh wagon tends to be more accurate than data from a grain cart or yield monitor. Published, peer reviewed studies tend to have the highest reliability. Third-party published data with statistical analysis and multiple sites years of data can also be highly valuable.