Question on chi2 test

[Vanter Birad]

Hi,

When I apply chi2- goodnes-of-fit-test STATA gives me the pearsons chi-square value of e.g. 10. There are 8 degrees of freedom. As a result 10 < 15.51 and H0 should be retained. My question now is which hypothesis STATA tests.

Is it Ho: F0 = F1 and H1: F0 != F1 or the other way around?

I guess it is the first one, but i am not really sure and can not interprete my results with not knowing that...

[ChR]

It is a goodness-of-fit model, so it tests whether one model better fits your data than the other. Hope that helps.

[Nick Cox]

Please study the Advice guide, especially sections 12 and 18. There is no context here of which command you used. But in general a chi-square test has a null hypothesis of equality.

[Vanter Birad]

Yes, that is what i know. My question is the following (maybe i need to explain it better).

Assume the dataset distribution is F0, my theoretical distribution is F1. I want to test if those two are the same or similar. Therefore i need to state hypothesis. Usually one would formulate (alternative hypothesis) H1 as: F0 is not similar to F1 (F0!=F1)

I am just wondering if STATA does test that way too....

[Nick Cox]

Stata (please do read section 18) is orthodox in how it sets up chi-square tests and your thinking is consistent with its formulation.

I can't add informative detail because you are still leaving the question vague.

[Vanter Birad]

Thank you Nick.
The question is vague because I have no specific example as of yet, I am in preparation of chi-squared test on my dataset and wanted to know, if my form of hypothesis formulation is right. If Stata adheres to H0: F0 = F1 and H1: F0 != F1, then I can retain null-hypothesis if the estimated value is smaller than the critical value and that is basically what I needed!

By Advice Guide you mean users manual? Thx and have a good sunday

[Nick Cox]

http://www.statalist.org/forums/help#gfaq_postingadvice

[David Hoaglin]

The original posting mentioned 8 degrees of freedom. What is the source of those degrees of freedom? Please tell us more about your data, and show the actual command that you used, so that we can try to give more-constructive comments.

[Vanter Birad]

The original posting mentioned 8 degrees of freedom. What is the source of those degrees of freedom? Please tell us more about your data, and show the actual command that you used, so that we can try to give more-constructive comments.

This is just a plain example from my university textbook...

Edit: So Stata, not STATA I'll remember it!

[David Hoaglin]

Even for a textbook example, it would help us to see the details, especially if the example resembles your data.

It seems that you plan to compare the distribution of a set of data against a theoretical distribution. A chi-squared test is one approach, but not the only possibility. You have not shared any information about the nature of your data, so it is difficult to give advice.