Trying to fully understand CHITEST function (or Goodness of Fit problem)

[zennie]

Hello.
Hope it's the right place to post such question.

Recently I have been dealing with logistic/sigmoidal function fitting with quite good results, thanks to this forum.
But now I am kinda stuck trying to figure the best way to use Excel to calculate how much two functions I have created fit each other.

I have created a simple two columns file, each column describes a certain sigmoidal curve (attached).
By plotting these curves together, one can obviously see they do not fit each other quite so good (although not VERY different).

When trying to measure the goodness of fit between these two functions using CHITEST, I come out with a very good fit of .99999 which seems quite untrue considering the apparent differences between the functions.

I'm beginning to think I have some major flaw in understanding the correct\appropriate use of CHITEST function. Maybe i should use a whole new way to assess the fit between these two columns?



Thanks so much for any reply.

[MarvinP]

Hi zennie and welcome to the forum,

On this page it says...
http://www.excelfunctions.net/Excel-...-Function.html

You should bear in mind that the chi-square test is not reliable when the expected values are too small. As a guideline, if any of the expected values are less than 5, or if the total of the expected values is less than 50, you should not rely on the result of the chi-square test.

[zennie]

Thanks Marvin.

A strange thing happens when I just multiply all scores by 100.

Now the CHITEST seems to provide a completely different result, saying the two distributions are totally different (p=.00000057).

Any reason for such GOF score? the two distributions are somewhat different but i would expect a much much higher goodness of fit score.

Thanks again.