Chi square on a huge dataset

AlexanderC

New Member
I have a situation where I want to look if a simulation gives expected values, so I think a chi squared is the way to do this. Some of the expected probabilities are quite tiny, but others are very large. After doing a chi square I noticed it gives a huge value = 867551.351! I see online that chi square is depending lots on the sample size so of course the p value is tiny and meaningless. My question is what is the correct way to look if observed and expected differ significantly, when we have such large sample and a reference distribution? Could I maybe do a log transform of the data?

Here is an example of the data which shows my problem.
Code:
class    ob    ex
1    0    10.78
2    98    97.3
3    1610    1680.7
4    10017    10087
5    14224    13755
6    181083    27475
7    133413    147896
8    301406    332773
9    2844688    2957983
10    3513461    3508239
If it is really problematic then I am not so interested in significance but would be nice to get indications of if there is a difference between observed and expected, and also WHERE are their differences.

Thanks

hlsmith

Less is more. Stay pure. Stay poor.
So where are these data coming from. How are you simulating them and the source of the reference?

Are they supposed to be counts?