Minoan is an IE language?

David Sánchez davius_sanctex at terra.es
Sat Mar 10 12:55:14 UTC 2001

> Wouldn't the question of sample size be relevant here? If inscriptions are
> very short, say only four or five words, wouldn't this make the chi-2 test
> invalid? As far as I remember, you would have to assume normalcy in order
> to use it, otherwise there are other tests with bigger error margins to be
> used. Could you define how big a continuous text sample must be before you
> can use the chi-2 test?

1. The size of the sample is important, all signs must appear at least 4 or 5
times, the total number ideally would be of 50 or more signs!

2. The test in the case of very short inscription is not invalid but in general
will fail in refute the "fundamental hypothesis", that is in general the test
will not say us more than we have assumed as "fundamental hypothesis". In
certain sense we can say that the test will be honest and if the data are very
scace it will be silent, but it will not lie!

3. Normality is not required to use a chi-2 test, that is precisely the strong
point of this test (it is a non parametrical test).

David Sanchez Molina (www.upc.es)
david.sanchez-molina at upc.es

More information about the Indo-european mailing list