Delving Into The Data

Are you confused on how to interpret the data coming off of your corn and soybean strip plots? The challenging 2007 season often impacted field variation and your test plots. Big differences in variety response have been observed, but how real are they? If they are real, we want to be choosing those superior varieties. This interpretation can be very difficult if no check plots are present in the individual tests. But even when check plots are included, it can be confusing to determine performance differences between the entries.

Consider a standard strip trial setup (Figure 1). Checks included in typical strip trial plots help to determine the amount of field variation present at the site. Site variation is a "killer" to see the true treatment differences that may exist. In this trial, we want to determine if there are real differences in yield potential between the varieties. Site variation can hide these differences, or lead us to incorrect interpretations.

Table 1 describes some possible situations that you might face with yield data this season. All 4 examples have a check average yield of 130 bu/ac. If we consider that a 10% variation from average of individual check yields is acceptable, then the range of check yields that would be acceptable is 117 to 143 bu/ac. If the check yields are within this range we will consider the differences between variety entries to be real.

¹ Calcualtion of Check Averages = Check 1A+1B+1C÷3

² Calculation of range around average yield: 130*0.9=117 and 130*1.1=143

Acceptable Field Variation

In Examples 1 and 2, all the check plots are within the 10% range and we have confidence that the field variation is minimal. We can use this plot data to choose varieties for next season.

Unacceptable Field Variation

Examples 3 and 4 highlight many of the situations occurring in 2007. The variation between checks is much greater than 10% of the check average. This suggests that other field related effects are having a significant impact on yield performance. These effects could be related to soil differences, including texture, drainage, compaction, etc. Past differences in tillage, fertilizer use, previous crop, weed pressure and other management practices could also be involved. In many cases in 2007 these field effects resulted in different parts of the same field having different availability to moisture.

Using plot data for choosing the best variety in these situations is nearly impossible. The only option left is to split the trial in two, using the central check plot and each outside check plot to determine an average for ½ the trial. If the individual two checks fall within 10% variation from the average of the two checks, you can consider comparing the varieties between these two checks. Even this does not work in examples 3 and 4. The variation in the individual checks from the average of the two checks is still greater than 10% (ie example 4 check 1A+1B = 100+128÷2 = 114 for a 10% range of 114+/- 10% = 102.6 to 125.4).

Examples 3 and 4 show trials that should not be used for making decisions on variety selection. Unfortunately, these trials are a lot of good intentions and effort that end up being wasted. 

Multiple Site Locations Required

Seed companies try to account for field variation by conducting numerous trials across the province, over a wide variety of locations. These trials primarily support the environmental adaptability of cultivars to a wide range of soil, climate and management effects. Previous studies have shown that a minimum of 30 site comparisons between varieties is required to offset the impact of field variation. Many of these strip tests are conducted without internal checks as described above. These trials must not be used alone in making variety decisions!

| Top of Page |