Yes. If you calculate the residuals (fo - fe)/sqrt(fe) [that is, the
square root of the contribution of the cell to the chi square statistic)
then on the null hypothesis that fo = fe, these are normally distributed
with mean zero and standard deviation given by
sqrt((1-rowtotal/grandtotal)*(1-coltotal/grandtotal)). You can then do a
simple Z test.

This of course assumes the expected cell frequencies are large
enough......

This set of tests corresponds to testing the multiple dummy variables
involved. I find it very useful - it makes the nature of the
relationship between the variables much more precise.

Regards,
Alan


Gerald Kaminski wrote:
> 
> Hello, All:
> 
> A general question concerning Chi-squared independence results.
> 
> In AOV, when there is significance, we can do pair-wise comparisons,
> build
> confidence intervals, etc.  Is there a comparable technique for
> Chi-square
> independence tests other than "eye-balling" the contribution to the
> Chi-square
> stat from each cell and looking for large changes?
> 
> Thanks,
> 
> Gerry
> 
> .
> .
> =================================================================
> Instructions for joining and leaving this list, remarks about the
> problem of INAPPROPRIATE MESSAGES, and archives are available at:
> .                  http://jse.stat.ncsu.edu/                    .
> =================================================================

-- 
Alan McLean ([EMAIL PROTECTED])
Department of Econometrics and Business Statistics
Monash University, Caulfield Campus, Melbourne
Tel:  +61 03 9903 2102    Fax: +61 03 9903 2007

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to