Cramér's 
is a measure of association between two categorical variables based on a contingency
table. For a table with 
rows, 
columns, sample size 
,
and Pearson chi-squared statistic 
, it is
 |
The normalization gives 
. A value of 0 corresponds
to no observed association, while a value of 1 is attainable for complete association
in suitable tables.
For a 
table, Cramér's 
equals the absolute value of the phi coefficient. As with
other effect sizes, its interpretation depends on the table dimensions and application.
