Agriculture Reference
In-Depth Information
the correct tolerance must be computed. Table 13C gives the “irst” and the “second” analysis and all appar-
ent differences. For a desirable component, the apparent deiciency is the value in column B minus that in
column C; whereas for an undesirable component, the apparent difference is the value in column C minus
that in column B. To simplify the testing, any component for which an apparent deiciency is not to be
tested may be omitted. Otherwise tolerances might be computed for weed seed, other crop seed, and inert
matter. Detailed calculations are provided in chapter 13 of the AOSA rules.
When weed seed and other crop seed are omitted, it is because their particle weights are not easily
determined. If tolerances are to be applied for any such components whose particle weight is unknown,
their particle weight must irst be determined. This is done by counting and weighing all the particles of
the component in the working sample or in a subsample of the working sample obtained with a mechanical
divider. Otherwise, computations are similar to those in Table 13C.
ISTA Purity Tolerances
The International Seed Testing Association has the following purity tolerances for different circumstances:
1. For purity tests conducted on the same submitted sample in the same laboratory (two-way test at
5% signiicance level), tolerances are shown in Table 3.1, ISTA rules, 2010.
2. Purity tests on two different submitted samples from the same lot when a second test is made in
the same or a different laboratory (one-way test at 1% signiicance level), tolerances are shown in
Table 3.2, ISTA rules, 2010.
3. For purity tests on two different submitted samples from the same seed lot when a second test is
made in the same or a different laboratory (two-way test at 1% signiicance level), tolerances are
shown in Table 3.3, ISTA rules, 2010.
These tolerances are all based on Tables P11, P1, and P7, respectively, developed by Miles in his 1963
handbook. The tolerances may be used for different purposes, depending on the situation and law enforce-
ment philosophy. The special tolerances developed by Miles for mixtures of species with different particle
weights which are in the AOSA rules have not been adopted by the ISTA rules.
ToLErAnCES for InCIdEnTAL SEEd ConTAMInATIon WITH noxIouS WEEdS
Statistical Assumptions
The occurrence of noxious weed seeds (or incidental seed contaminants) and their variability in repeated
samples taken from the same population under certain assumptions can be described through the Poisson
distribution (Collins, 1929; Leggatt, 1935a, 1935b; Miles, 1963; Elias et al., 2000). For the application of
tolerances based on this distribution, it must be assumed that the occurrence of such contamination in each
sample is independent of that in any other sample and that the population mean does not change following
its estimation for labeling.
At low population means, the Poisson distribution is nonsymmetric with a positive skew (Fig. 13.2).
As the population mean (e.g., the true number of noxious weed seeds or incidental contamination) increases,
the distribution becomes nearly symmetric and approaches the normal distribution. Thus, at higher popula-
tion means (i.e., higher levels of contamination), a normal approximation of the Poisson distribution may
be used for development of tolerances.
Under the Poisson assumption, the mean (μ) of a population must equal its variance (σ
2
). However,
since in practice these parameters are generally unknown, their estimates based on the number of noxious
weed seeds labeled or represented in a submitted sample are used (Hahn and Chandra, 1981). Under this
