Agriculture Reference
In-Depth Information
13
Statistical
Applications to
Seed Testing
SECTIon onE - ToLErAnCES
Theory and Application
This section will explain the types of error that may occur in making a decision to accept or reject sam-
ples that represent seed lots. It also will discuss the sources of variation that may describe the differences
between accuracy and precision, and the use of tolerance tables in seed testing.
It is well-known that repeated tests on samples of the same seed lot do not necessarily produce identi-
cal results and that test results, even on the same sample, can be expected to vary to some extent. Variability
in seed testing is accounted for through the use of tolerances which specify the limits by which repeated
independent test results can differ and still be considered consistent with the labeled value. They allow seed
regulatory oficials to determine if seed on the market conforms to truth-in-labeling regulations and provide
a mechanism for consumer protection in the commercial sale of seed. However, they also protect seed sup-
pliers by deining the limits by which quality tests can vary from the labeled rates and still not be considered
out of tolerance. Thus, they are an important factor in the orderly marketing of seed.
In the enforcement of labeling laws, seed regulatory oficials routinely obtain representative samples
of seed from points of sale and submit them for a law enforcement testing. Depending on the results of such
oficial tests, determinations are made about whether lots are correctly labeled, or whether there is a
sig-
niicant
difference between the labeled quality and that of the second test. As in other tests for signiicance,
it is expected that a correct decision will be made 95% of the time, assuming a 5% probability level. The
irst part of this chapter will explain types of errors that may occur in making a decision to accept or reject
samples that represent seed lots. It also will discuss the sources of variation that may describe the differ-
ences between accuracy and precision, and the use of tolerance tables in seed testing.
Types of Error in Enforcing Labeling requirements
If a seed lot is considered to be incorrectly labeled (germination, purity or other), when in fact it is not,
a Type-I error (producers risk) is made. This type of “false positive” error might be made by an oficial
inspector who determines that the germination of a seed lot is out of tolerance with the labeled level when
in fact it is not. Thus, Type-I errors work against the interests of seed producers who ind the sale of their
seed in jeopardy. On the other hand, Type-II errors (consumers risk) may be made by a seed inspector who
determines that a seed lot is correctly labeled when in fact it is not. Thus, Type-II “false negative” errors
work against the interests of the consumers who may be hurt by such errors because they can result in the
sale of poor quality seed. Generally, law enforcement decisions concerning potential labeling violations
are made only after additional tests are completed to conirm or reject preliminary results. Thus, incorrect
decisions can often be avoided by care in enforcement decisions.
293
