As we all know I can get very confused by statistics. What, for example, is a representative sample? EU Rota asks the same question.
Although not a trained statistician, it would appear to be absurd to
base the regulatory apparatus of an entire continent on 0.000000086% of
the total population.
Anyone care to tell me whether this is a vlaid sample or not?
In
I’d say that if we decided that any of these were possibly harmful, then it’s enough evidence to spend the money on a larger survey that would be necessary for a quantitative result.
I don’t see what difference it would make to the regulatory process whether 100% or 0.1% of the population were contaminated.
Small samples can yield significant results (or rather it is correct to say that large samples do not in and of themselves confer greater statistical significance unless attention is paid to possible confounding factors). But 39 families—150 people, say— is going to have a standard error of the order of 8% That will wash out most statistics, especially the lamentably weak ones typically used in public health studies.
If your sample is representative, then there’s no need to relate sample size to population size. If the sample is not representative, then the whole thing may be rubbish anyway, to some unknown extent. Much of stats seems to consist of the rote application of simple little formulae to data whose validity is doubtful. For what it’s worth, I’ve found professional statisticians pretty helpful on this sort of thing, and pretty frustrated at the sort of oiks who abuse statistical procedures or results: journalists, medics, those sorts of people.
Leave a Reply