Section 2 - METHODS

Statistical basis of sampling

We often use very small samples to infer about the condition of a large flock of birds. Appropriate statistical techniques are available to determine the number of samples necessary for a given purpose. It is often the economic and practical considerations which actually determine the number of samples taken. Sampling should generally be stratified. \par \par Stratification is the drawing of animals to be sampled from different groups , normally in proportion to the total number of animals and of the degree of observed variability within each group. If, for example, we want to estimate the situation in a whole poultry farm or nucleus, birds should be sampled in each of the houses, and not merely in the first division of the first house! Proportionately more samples would be drawn from age groups which past experience has shown to have more highly variable results.

On the other hand deliberately biased sampling is sometimes necessary , and even desirable, if we are investigating a specific problem. If the problem is paralysis in breeding hens we should not base our opinion only on examination of birds showing no symptoms!

The actual number of samples needed will vary with the number of birds in the flock(s), within-flock variability of response, division and separation of groups, type of tests to be run, the objective (routine monitoring vs. diagnosis of a problem), and the nature of the disease involved (rate of spread within and between houses). In practice 12 to 24 samples per air space the typical range for most tests. Sixty samples per air space is recommended where the objective is to detect infection with an unwanted pathogen (e.g. Mycoplasma sp. or S.enteritidis).

If within-flock variability is low or unimportant, it may be possible to run the tests on pooled samples. It must be realized that using pooled samples can reduce the sensitivity of the tests due to positive samples being diluted out by the negatives. In any case pooled samples should be properly constituted with equal quantities of the component samples. Wherever possible the individual samples should be submitted and the "pooling" performed in the laboratory which will conduct the tests. In this way the laboratory will always have the option to repeat the test on the individual samples when in doubt.

Sampling strategy is further discussed in the section on Clinical Interpretation (p.30)