Re: Questions on the F-Test *VS* the Chi-Square Test

Date: Nov 24, 2005 5:23 AM





David Jones said:


***Well, what I meant by a chi-squared test was that deriving directly
from the theory of Normal distributions, where you have a sum of squared errors, or a reduction in sum of squared errors, which follows a chi-squared distribution multiplied by the true variance, leading to the chi-squred test for whether a given fixed value for the variance can be accepted, or to a confidence interval for the variance.
It may be that you are thinking of a chi-squared test in the context of a contingency table, where things are a little different (and the chi-squared distribution is only approximate). Still such tests can often be viewed as a comparison between the variation of counts about their exected values (so nearly equivalent to an estimated variance), compared with a known value (which may not be explicitly there, but derives from the sampling theory for the counts given the null hypothesis). David Jones***

My feedback:

Of course, I beg your pardon. I read hastily, then inaccurately. You are concerned with what is currently named a *Goodness-of – fitt* problem.

__________licas (Luis A. Afonso)
.