|
 Essay - july/aug 2007
Statistics, approximations and close truths III
James C. Delouche
Professor Emeritus Mississippi State University
Most countries have legislation that regulates the production and/or marketing of seeds to protect its farmers. Farmer needs protection against misrepresentations of the seeds they purchase for planting because their livelihood and in many cases essential food supplies for their families depend upon an adequate production. The well being of farmers and the rural economy are, of course, also vital national interests. Seed legislation differs among countries mainly in the degree of control over seed production. Both production and marketing are closely regulated in some countries while in others regulation is mostly confined to the marketing phase. Marketing regulations variously require that seeds offered in the market meet specified quality standards, be truthfully labeled for specified quality attributes, or both. Enforcement of the regulations is accomplished through inspections in the production site and marketplace and testing to determine that the seeds meet the specified quality standards and/or are truthfully labeled for the specified quality attributes. Seed inspections and testing, however, begin long before the seeds enter the marketplace. They are crucial activities in the seed producer's quality assurance program to ensure that the methods used to prepare seeds for marketing are appropriate, properly applied and effective.
Sampling
Sampling is the critical step in the application of the statistical procedures to determine or verify the quality of the huge populations of seeds in the original seed bulk and eventually the lot or batch of seeds available for purchase in the market. The relatively small sample, usually a kilogram or less, taken to establish the physical and varietal purity, germination, and other quality attributes must represent the multiple tons of seeds in the bulk bin or lot of seeds in packages from which it was taken. A good analogy is the several thousand person sample taken by political pollsters to determine the political preference of an electorate of multiple millions of voters. Sampling is the key step. If the sample is not representative the determinations based on it may be no better than guesses. Considering the importance of the sample in determining seed quality, it is not surprising that it has received and still receives much attention. The American Association of Seed Control Officials, for example, recently published a Handbook on Seed Sampling that thoroughly covers the subject from equipment and procedures, records, recommendations on seed lot size, safety and health issues for inspectors, to references and on-line resources.
There has been much progress in seed sampling especially for internal quality control purposes during preparatory operations such as drying, bulk storage, and processing. Automatic samplers strategically positioned in the various conveyors have largely replaced the hand-held "dippers" used to extract portions of seeds from the seed stream. Vacuum samplers permit much better and much, much easier sampling of seeds in bulk bins. Sampling of bulk cottonseed before delinting, for example, was a hot, arduous, exhausting and mostly hit-or-miss task before the development of a suitable vacuum probe sampler. It still is a hot task but not too arduous.
The sampling process
Sampling is not complete with the taking of primary samples and combining them into the composite sample. In either case the composite sample contains a lot more seeds than can be practically used for the various quality determinations. The composite sample, therefore, has to be divided, i.e., repeatedly sub-sampled, to obtain the working sample which is about 3000 seeds for the purity analyses and 10X that number for special examinations such as for the presence of seeds of noxious weeds. The various types of dividers used are precision mechanical devices that thoroughly blend the seeds in the composite sample and extract the portion of the seeds needed for the working sample. The 400 seeds then taken from the pure seeds for the germination test constitutes an even more minute, even infinitesimal, portion of the seed lot. Consider, for example, that in the case of a 10 ton wheat seed lot the 400 seeds tested for germination represent about 200 million seeds. Remarkably, however, when the proper equipment and procedures are used for the samplings the statistical results are very acceptable and repeatable approximations of the "true" values.
Non-uniformity of Seed Lots
The big problem in sampling is non-uniformity - or heterogeneity - of the bulk seeds or packaged seed lot sampled. If the seed lot is uniform throughout its parts for the quality factors appearing on the label as defined in many seed laws, a single primary sample of the size required for the purity and special analyses would be sufficient to establish or verify quality. Most seed lots, however, are not that uniform so primary or sub-samples have to be taken from various portions of the bulk or bagged seeds. For official inspections a primary sample is taken from 10% of the number of bags in the lot or the equivalent in bulk seeds but not more than 30. Even with these multiple primary samples it is frequently not possible to obtain valid, repeatable results from very non-uniform seed lots. Furthermore, the uniformity or non-uniformity of a batch or lot of seeds often changes during processing and conveying of the seeds due to mixing and stratification as discussed by M. Kruse in his article on Heterogeneity in Cereal Seed Lots in the July/August, 2005 issue of SEED News.
The causes of non-uniform seed lots have been previously noted: non-uniform production fields, mixing of different batches of seeds in bulk storage, poorly adjusted equipment, inadequate processing and so on. The main remedy for non-uniformity is blending, but intra-lot blending, i.e., the mixing of a non-uniform seed lot to acceptable uniformity, can be difficult to impractical. Lot size usually has to be greatly reduced to permit the kind of agitation and mixing needed to achieve the desired uniformity. On the other hand, inter-lot blending, i.e., mixing of two or more uniform seed lots, to produce a new lot is a much simpler and more frequently used procedure. Seed lots of lawn and pasture seed mixtures of as many as six species are routinely produced by careful calculation of the proportions needed, accurate metering of the seeds from the different batches of seeds and some type of in stream mixing device. Similarly, two or more lots can be blended to "dilute" the presence of noxious or other undesirable seeds or to salvage low germination seeds by mixing them with high germination seeds. Non-uniformity of seed lots is a perennial and important problem that must be prevented or resolved as it causes unintentional and apparent misrepresentations of seed quality that attract the attention of seed control officials and results in dissatisfied customers.
Games Seedsmen Used to Play
The procedures used in the official sampling of seeds for regulatory control provide some opportunities for misrepresentations which unfortunately have been taken advantage of by unscrupulous persons. I recall several cases in which an attempt was make to profitably dispose of bags of poor quality seeds by mixing some of them with the bags in a good quality seed lot. The proportions of bags of low quality seeds and bags of good quality seeds and the labeled quality values were craftily calculated to produce a high probability that the primary inspection samples from the desired number of individual bags when combined into the composite sample would produce test values within tolerance of the labeled values. We termed this devious procedure "bag blending" because the inspector sampling the seeds effectively "blended" the seeds from the poor quality bags with those from the good quality bags. When "bag blending" was suspected, we had the inspector take and maintain the identity of samples from the individual bags for testing to determine the variation in quality between bags. There were other devious manipulations used by unscrupulous persons in attempts to defeat the sampling protocols. One frequently used manipulation in the case of seed lots that contained prohibited or restricted noxious weed seeds was to start off with a relatively large lot size and if noxious weed seeds were found to divide the lot into two or more new lots for retesting with the hope that noxious weed seeds would not be found in at least one of them. We termed this procedure "divide and conquer." Fortunately, devious manipulations and frauds of these sorts have largely disappeared with the professionalism of the seed industry, the sophistication of the farmers, and in the United States, at least, an abundance of lawyers.
Inspections and sampling involve mostly statistics
Determining the important quality attributes of the seeds sampled is accomplished by testing which also involves statistics and more than a few approximations and close truths. These will be considered in the next and final part of this long essay on Statistics, Approximations and Close Truths.
|
|
............
|
