Value in Performance Trial Results

When looking at test plots everyone naturally looks at what was the highest yielding product in the plot. The question that should be asked is, “is it really?”

Beyond raw genetics there are some key factors in determining yield in a test plot. Just like real estate, it can be about ‘location, location, location’. Few test plots are perfectly flat, have completely uniform soil and the same moisture available. Sometimes you can see this in plots that have the same check variety planted every 2-4 slots throughout the plot. On one end it is yielding 150 bushels and by the time you reach the other end it is yielding 175 bushels. Same genetics; different yield.

This raises the question, “is the highest yielding variety truly the highest yielding product or is it just in a good spot?” That can be difficult to determine in small plots run by customers or dealers. The problem becomes even bigger when you get to State and F.I.R.S.T. trials when you have 50+ varieties planted over several acres. The likelihood of the highest yielding variety being genetically higher yielding than everything else in the test is very low.

The people running these trials recognize this issue and employ a strategy to show how much different environments within the test plot are affecting yield. The testing organizations that conduct these trials typically grow the entries at multiple locations with 2-3 replications of the varieties at each location. A replicated trial such as this enables the reader to develop a level of confidence when comparing the varieties in the trial’s results. If the same varieties appear in the top of each replication and across locations, then the confidence in comparing the varieties is very high. The more the varieties jump around in the order relative to each other, the less confidence there is in determining if one variety is actually higher yielding than another.

But what if we do not see the individual replication or location data? How do we know if one variety is better than another with any degree of confidence?  All replicated trials should have a statistic at the bottom of the table that will help address this question. That statistic is called “L.S.D.” or Least Significant Difference. If the difference between two varieties is less than or equal to the L.S.D. value, then the two varieties are not statistically different.

The L.S.D. value is calculated by determining to what degree the relative rank order of the entries is jumping around from one replication to another and from one location to another. The larger the L.S.D., the more the rank orders of the varieties have changed relative to each other going across replications and locations. In other words, it is a statistical calculation that gives the reader a way to compare, with some degree of confidence, two varieties in a table of multi-location averages. If the L.S.D. for grain yield in a corn trial is 12 bushels, then the reader can assume that unless any two varieties differ by more than 12 bushels, their performance is not statistically different. I have seen some performance trial results from this year where the L.S.D. value is greater than the difference between the highest yielding variety and the lowest yielding variety; no significant difference among all varieties in the trial.

L.S.D. essentially shows how much environmental variation was in the plot and any product that has yield equal to highest yielding variety minus L.S.D. should be considered to have the same genetic yield for that test. In our company we term these as ‘Top Finishers’. So, be careful when you look at performance summaries. Make certain that the first value you look at is the L.S.D. It is the most important one.