distribution of confidence intervals. Modified stability
analysis is based on a procedure long used by plant
breeders. It is a means of using the variability found on
different farms to help confirm recommendation domains or
indicate a need to partition the domain. The modified
stability analysis is a relatively simple procedure which
utilizes an environmental index created by on-farm trial
data as a means of measuring all the factors which
influence response to a technology. These factors include
climate, soils, and farmer management practices. A simple
associated procedure, a graphic distribution of confidence
intervals, is used to evaluate the variability in results
to be expected from a technology within a recommendation
domain. Both procedures are amenable to calculation with
simple calculators; neither requires, though both can
benefit from, replicated designs. They are specifically
useful for on-farm trials conducted at several locations.


Modified stability analysis

 To understand modified stability analysis
(Hildebrand, 1984), consider farmer-managed trials
conducted over a large number of farms within one
preliminary recommendation domain and utilizing an
improved cultivar compared with a local variety. No other
changes are made from farmers' usual practices. The only
constant at each location (farm) is the cultivars. Each
farmer subjects them to different soil conditions,
planting dates, pest control, fertilizer, and other
management practices. A farm for which the average yields
of the two cultivars is high (for whatever reason) is
considered to be a "good" environment for the crop as
measured by the average yield. A farm for which yields
are low (for whatever reason) is considered to be a "poor"
environment. Environment, then, becomes a continuous,
quantifiable variable whose range is the range of average
yields from the trial. Yield for each of the varieties
can be related to environment by simple linear regression,
based on the following equation: