Fig. 5.

Asymptotic variances of correlation estimators when the true underlying distribution is skew-normal. The true correlation coefficient is 0.4. a ${\mathit{\alpha }}_1=0.1$ and both ordinal variables have five categories. b ${\mathit{\alpha }}_1=0.5$ and both ordinal variables have five categories. c ${\mathit{\alpha }}_1=1$ and both ordinal variables have five categories. d ${\mathit{\alpha }}_1=0.1$ and both ordinal variables have three categories. e ${\mathit{\alpha }}_1=0.5$ and both ordinal variables have three categories. f ${\mathit{\alpha }}_1=1$ and both ordinal variables have three categories.