# Cramer's V in XLSTAT

#### Austringer

##### New Member
Hi all,

I performed contingency table tests and I want to know how to interpret Cramer's V test stat. I am using Cramer's V as my tables violate basic Chi-square assumptions of theoretical frequencies >5, i.e. my freqs. <5 in tables such as below. Which p-value can I use to interpret the Cramer's V test in XLSTAT if I set significance level to 5%? I know how to interpret the V-value.

Test of independence between the rows and the columns (Chi-square):
Chi-square (Observed value) 1.011
Chi-square (Critical value) 7.815
DF 3
p-value 0.799
alpha 0.05

Cramer's V = 0.100

Theoretical freqs.
Obligate Facultative Total
Urban 3.088 31.912 35.000
Golf 2.559 26.441 29.000
Farm 1.941 20.059 22.000
Control 1.412 14.588 16.000
Total 9 93 102

#### gianmarco

##### TS Contributor
Hello!
The chi-square test is actually informing you that there is NOT a significant association between the categorical variables which make up the table. As you can see from the results, the observed chi-sq value is less than the critical value (for 3 DF). The associated p value is far above 0.05. That said, it is useless using Cramer V as a coefficient which quantifies the strength of the correlation, since the chi-sq test has already told you that there is not correlation.

If you had got a significant chi-sq result, then you could have gone on trying to measure that correlation via Cramer V (or other measure of categorical association).

Hope it helps
Best
Gm

#### Austringer

##### New Member
Also note that the theoretical/expected freqs. have some values below 5.

What test of association should I use then if Cramer's V and Chi-square is rendered useless here?

Regards,

#### gianmarco

##### TS Contributor
Hello!
I do not know how one could "fix" the issue of the presence of expected values <5.
I am unsure if re-sampling techniques could help in this situation....

I would like to know the opinion of other users
Gm

#### Austringer

##### New Member
Hello!
I do not know how one could "fix" the issue of the presence of expected values <5.
I am unsure if re-sampling techniques could help in this situation....

I would like to know the opinion of other users
Gm