[R] chisq warning

Dieter Menne dieter.menne at menne-biomed.de
Fri Aug 12 14:37:03 CEST 2005


Stephen Choularton <mail <at> bymouth.com> writes:

> I am running chisq as below and getting a warning. Can anyone tell me
> the significance or the warning?
> 
> > chisq.test(c(10 ,4 ,2 ,6 ,5 ,3 ,4 ,4 ,6 ,3 ,2 ,2 ,2 ,4 ,7 ,10 ,0 ,6
> ,19 ,3 ,2 ,7 ,2 ,2 ,2 ,1 ,32 ,2 ,3 ,10 ,1 ,3 ,9 ,4 ,10 ,2 ,2 ,4 ,5 ,7 ,6

.....

> Warning message: 
> Chi-squared approximation may be incorrect in: chisq.test(c(10, 4, 2, 6,
> 5, 3, 4, 4, 6, 3, 2, 2, 2, 4, 7, 10,  

This warning means that the expected count in one group is < 5. Try 
simulate.p.value = TRUE if you want to be sure your result is meaningful. 
Looks like you should be cautios: p = 0.0004998 versus p< 2.2 E-16 with
simulate.p.value = FALSE.

Dieter




More information about the R-help mailing list