Transcript Document

Precision Farming Profitability
Reference G: On-Farm Research
By Sylvie Brouder, Robert Nielsen
Replicate
number
Note: Cannot tell if yields reflect something
about the treatment, something about the
field gradient or some about both factors.
Treatment A
Treatment B
Note: Completely random assignment of
treatments may still result in bias. Here,
purely by chance, more of the A treatments
are upslope from B Treatments.
Treatment A
A
B
Treatment B
Rep.1
Rep.2
C
Rep.3
Rep.4
Treatment A
Treatment B
For most trials, randomization of
treatments within each replicate is
the best option.
Figure 1a. Side-by-side comparison: no replication. B. Completely randomized
design w/4 replicates. C. Randomized block design w/4 replicates.
Precision Farming Profitability, p. 104
Case 1: No difference in yields.
Case 2: No difference in average
yields but replicates vary.
Figure 2. Three cases where statistics are not needed to interpret the results
from a 2 treatment trial with 4 replicates.
Precision Farming Profitability, p. 106
Case 3: Treatment B out-yields
Treatment A in every replicate.
Case 4: Treatment B yields more
than Treatment A in 3 replicates but
yields are the same in one replicate.
Case 5: Treatment B out-yields A
in 3 replicates but A out-yields B
in 1 replicate.
Figure 3. Three cases where statistics are most likely needed to correctly
interpret the results from a 2 treatment trial with 4 replicates.
Precision Farming Profitability, p. 107
Case 6: Results in individual replicates
are mixed but Treatment B averages 5
bu/acre more than treatment A.
Table 2. Understanding the probability.
Precision Farming Profitability, p. 108
Table 3. In each case average yields in Treatment B are greater than A. Test your
statistical IQ by figuring the odds that the Treatment B cases are really no
different than Treatment A. In cases 6-10 look at only the first 3 replicates of
Treatment A.
Precision Farming Profitability, p. 109
Figure 4. Design for an on-farm trial comparing 2 treatments using 4 replicates.
Treatments are randomized within each replicate.
Precision Farming Profitability, p. 110
Replicates
1
2
3
4
1
Treatment A
Treatment B
Treatment C
Figure 5. Design for an on-farm trial comparing 3 treatments using 4 replicates.
Treatments are randomized withi9n each replicate, and replicates are located so
as to avoid known areas of extreme variability.
Precision Farming Profitability, p. 110
Figure 6. Design for an on-farm trial using the “split-planter” approach. This
layout shows the equipment passes when planter and combine widths are the same.
Precision Farming Profitability, p. 111