Cramér's V — Definition, Formula & Example
Cramér's V is a number from 0 to 1 that tells you how strongly two categorical variables are related in a two-way table. A value near 0 means almost no association, and a value near 1 means one variable nearly predicts the other.
For an contingency table with observations and Pearson chi-square statistic , Cramér's V is . It satisfies . exactly when every observed count equals its expected count under independence. when the association is perfect. For a table, equals the absolute value of the phi coefficient.
Key Formula
Where:
- = Cramér's V, the strength of association, between 0 and 1
- = the chi-square statistic computed from the table
- = the total number of observations in the table
- = the number of rows (categories of one variable)
- = the number of columns (categories of the other variable)
Worked Example
Problem: A school surveys 250 students. It records each student's grade (10th, 11th, or 12th) and favorite lunch option (pizza, tacos, salad, or sandwiches). This gives a 3-row by 4-column table, and a chi-square test on it gives . Find Cramér's V.
Step 1: Find the smaller of (rows minus 1) and (columns minus 1). The table has 3 rows and 4 columns.
Step 2: Multiply the sample size by that value to get the denominator.
Step 3: Divide the chi-square statistic by the denominator and take the square root.
Answer: Cramér's V is 0.2. Grade level and lunch preference show a weak-to-moderate association.
Why It Matters
A chi-square test in AP Statistics tests for evidence of an association. It does not tell you how strong the association is, and with a large sample even a tiny association can give a very small P-value. Cramér's V reports the size of the effect on a fixed 0-to-1 scale, much as r does for two quantitative variables. Cramér's V itself is not on the AP exam, but survey researchers and social scientists report it routinely.
Common Mistakes
Mistake: Treating a large Cramér's V as proof that the association is statistically significant, or a small V as proof that there is no association.
Correction: V measures strength, not significance. Use the chi-square test's P-value to judge significance and V to judge size. Report both. Also check the denominator: it uses the smaller of (r − 1) and (c − 1), not r or c themselves.
