How to test chi-squared dist?

How can I test whether my observed values come from a chi-square distribution with n-degrees of freedom (preferably with Anderson Darling in Matlab)? Thanks.

Answer 1

[Star Strider]

That does not seem to be an option, even in the Wikipedia article on the Anderson-Darling Test. I’ll be certain to be corrected here if I’m in error, but to the best of my knowledge, statistics (derived from data) can be described by the Chi-squared distribution, but data itself cannot be (or at least to the best of my knowledge, have not been described to be).

You can possibly do a nonlinear curve-fit of the chi-squared PDF (or CDF) to a histogram (or a cumsum histogram) of your data, but I doubt it would be very informative.

[Fred]

Yes you are wright I checked it again and it seems these so called observations are statistics!. So how can we check these statistics if they can be described by a Chi-squared distribution?

[Star Strider]

I would simply do the nonlinear curve fit I described in my initial Answer, preferably using nlinfit or fitnlm (which is a bit easier to use and gives you all the statistics you need on the fit without your having to do anything further). However I still do not see what useful information you would gain from doing that.