Figure 7.

Tuning probabilities and speedup for Scenario 2 datasets. The proposal probabilities used in Algorithm 1 for the root were summarized for the Scenario 2 datasets. With the true model containing an interaction between X₇ and X₁₀, the performance was dependent on both the graph used and where the variables were mapped to the graph. The symmetric graph with the causal variants mapped to the same concept had the best performance. Note that the whiskers and outliers were not drawn on the box plots.