## Chi-square value

### Chi-square value

Hi I was just wondering if there is a minimum required value for
chi-square or is it enough to have sig. under 0.05 and less than 20%
of the cells with values under 5 for the test to be significant? I've
heard somewhere that the value should be over 7 or am I wrong?

Elin Sigurdardottir

### Chi-square value

> Hi I was just wondering if there is a minimum required value for
> chi-square or is it enough to have sig. under 0.05 and less than 20%
> of the cells with values under 5 for the test to be significant? I've
> heard somewhere that the value should be over 7 or am I wrong?

> Elin Sigurdardottir

That the minimum value of the chi-square test statistic is
7, you mean?  I've never heard anything like that.  Here is
a summary of the conditions under which the appropriate
chi-square distribution provides a good approximation to the
sampling distribution of the test statistic:

http://www.angelfire.com/wv/bwhomedir/notes/chi_square_assumptions.txt

Cheers,
Bruce
--
Bruce Weaver

www.angelfire.com/wv/bwhomedir/

### Chi-square value

The acceptable value depends upon your degrees of freedom.  You can cross
reference this with a chi-square table.  Not sure where you could find one
online.

Depending on your degrees of freedom 7 could be low or high.

If you have significance of 95% or greater, then you're (generally speaking)
OK, regardless of the actual chi-square number.

Tony

Quote:> Hi I was just wondering if there is a minimum required value for
> chi-square or is it enough to have sig. under 0.05 and less than 20%
> of the cells with values under 5 for the test to be significant? I've
> heard somewhere that the value should be over 7 or am I wrong?

> Elin Sigurdardottir

### Chi-square value

I must be confusing it with something else I thougt so actually but i
wanted to be sure ;)

Elin

An analysis of survey data obtained using a stratified sampling design
gives us Region (4 values) and alcohol measures such as Heavy drinker
(yes/no).  The chi-square analysis yields both LLCHISQ and CHISQ
estimates.  When is it appropriate to report one or the other of these
results?  Achieved levels of significance vary ... e.g.,
ChiSq = 6.74
p = .0811
LLChiSq = 36.00
p = .0000

Thank you,
Laurel & Kyunghee