Failing to reject a causal model does not exclude all other possible hypothesized causal models. One reason is insufficient statistical power. AIC statistics can be used to rank competing non-rejected hypothesized causal models. Equivalent models are alternative causal graphs that make the same d-separation claims and therefore cannot be distinguished even with perfect statistical power. There are simple rules for obtaining all equivalent models that make the same d-separation claims, and often partial independent knowledge of the causal direction of some direct cause–effect claims can exclude many of the alternative models.