Re: chi-squared test. hypothesis confusion

"mikeeria" <michel.kowalczyk@xxxxxxxxx> wrote in
news:1135348984.853544.285020@xxxxxxxxxxxxxxxxxxxxxxxxxxxx:
> My questions
> -Does "reject Ho" means that we accept H-alt ???
> -Does "cannot reject ho" means that we can accept H-o ???

Those are actually philosophical questions with no hard answers. In
practice, most researchers who are able to reject the null hypothesis will
regard it as reason to *tentatively* accept the alternate hypothesis.
People are more split on the second one, but a common position is that
"cannot reject null hypothesis" means "I haven't *yet* seen enough evidence
to make me change my mind."

In practice, how to answer the questions depends not only on your
philosophical mindset but also on the real-world implications of the test.
You have to understand how much damage finding a phony association would do
and balance it against the damage done by failing to find a real
association. These can only be determined by subject-matter
considerations, *not* by statistical ones. Decision theory provides a more
formal framework for making these tradeoffs, but it still requires subject-
matter specific information.

.

Quantcast