Fig. 3
From: A dependence detection heuristic in causal induction to handle nonbinary variables
Proportion of samples (i.e. contingency tables) for which pARIsmean (left column) or MI (centre column) can be calculated, and the difference between them, for sample sizes of 10 [(a), upper row], 100 [(b), middle row], and 1000 [(c), lower row].
