Sales, sales, sales! We are going to continue rolling through our sales series and talk a little bit about statistics. What do statistics have to do with agricultural sales? A lot more than you think! Statistics will show you if a product is even worth looking at. Also a little knowledge on statistics will help you sift through all the literature you are shown about how great some products are.
I have hanging on the wall of my office the first document I produced from my Master’s Thesis Research and I’ll be the first to tell you that it is misleading. I was researching how cover crops impacted winter annual weeds the following spring. I crunched the numbers real quick to calculate averages and found that in that season, in Kansas, cover crops had reduced winter annual weed density in the spring by at least 50% compared to fallow! Sounds pretty good? Well, it wasn’t, averages only tell you part of the story. I had a fallow plots that ranged from 12 to 150 weeds. The average was 50. A forage bean plot ranged from 0 to 100 weeds, the average was 25. Were they that different? Averages can look good in a graph or table but they can be misleading.
If someone tries to sell you a chemical, hybrid or other product claiming a yield increase, look at the data they give you. Is there a graph or text that shows a difference between products? Hybrid A may only yield an average of 155 bushels per acre (bu/A) of corn but Hybrid B may yield 175 bu/A. What you need to ask is if this difference is significant. Look for things like standard deviation or least significant difference (LSD). Standard deviation gives you a range around the average. It will tell you Hybrid A yielded 155 bu/A with a standard deviation of 20 bu/A. This means 75% of the test plots for Hybrid A averaged from 135 to 175 bu/A. You could expect the same variability in the field. The more the standard deviations overlap, the less I would trust the product. LSD is a measure of how different the averages need to be before they are considered statistically different. In this case Hybrids A and B might be listed with a LSD of 15. Hybrid B averages 20 bu/A more than Hybrid A. This difference is greater than the LSD of 15 bu/A so Hybrid B did in fact yield better than Hybrid A.
If you are handed a graph, look closely at it. Is there standard deviation or LSD listed? Look at the scale of the graph. I once saw a graph that at a glance showed a huge difference between two products only to find on closer inspection that the yield difference was only two bushels, a gain not worth the cost of the product. If you don’t see what you are looking for you can always ask. I’m a firm believer that there are no dumb questions, just dumb answers. Ask the salesman if the products did in fact have different results. If he doesn’t know, in order to make the sale he should be willing to search through his company to find the answer.
Another good question to ask is how the research was conducted. Was it a single strip in the field, a split field, or a randomized, replicated design? You want a randomized and replicated design. This is the only design that will minimize variability related to soils, nutrient levels, insect, weed and disease pressure in a field. It is also the only design that will allow a proper statistical analysis.
Before you commit your money to a product, make sure the claims the salesman are making are true. If they can’t provide the proper information, their product should be held suspect until they can. For help sifting through the endless data about various products, or for help in utilizing statistics to make your on farm research more meaningful, contact Justin Petrosino at the OSU Extension Office at 548-5215 or petrosino.3@osu.edu.
Justin Petrosino, ANR Extension Educator, OSU Extension, Darke County
No comments:
Post a Comment