chi2gof gives very different p values for the data sets with the same range.

I am trying to do the fitting test with chi2gof for the normal and lognormal distribution. My data set is an output obteind from a problem with 8 variables (with different distributions). On the graph obtaind with dfittool, the output seems to fit the lognormal distribution well.

After performing the chi2gof for distribution lognormal I obtained the p=0.60878. Than I prepared a new data set, in the same way before(the range of variables the same). New fitting showed p=0.0004. Every time I repeat this, the p value is very different and I can't find any reason why sometimes it shows very good fitting and sometimes reject it. I also tried with different amount of values, from 50 up to 30000.

Is there any logical reason for this type of behaviour of chi2gof test?

[Craig Feied]

Did you ever figure this out?

Matlab doesn't seem to be a very good place to get answers.