Cramer's V is a scaled version of the chi-squared test statistic \(\chi^2\) and
takes values in \([0, 1]\). It is calculated as
\(\sqrt{\chi^2 / (n \cdot (k - 1))}\), where \(n\) is the number of observations,
and \(k\) is the smaller of the number of levels of the two variables.
Yates continuity correction is never applied. So in the 2x2 case, if x is the
result of stats::chisq.test(), make sure no continuity correction was applied.
Otherwise, results can be inconsistent.