Table 3.

Friedman example: $\text{Brier score}=\frac{1}{P}{\sum }_{p=1}^P{(I_p-{\mathit{\text{VIS}}}_p)}^2$ where I_p = 1 for truly important variables x₁ … x₅ and I_p = 0 otherwise. Hence the lower the Brier score the better the model variable selection. Items in bold show the best model with respect to the Brier score.

p	BART BMA	RF	bart Machine	dbarts default	dbarts best
100	3.24 × 10−2	3.82 × 10−2	4.30 × 10−2	4.66 × 10−2	4.17 ×10−2
1000	3.26 × 10−3	4.00 × 10−3	4.62 × 10−3	4.87 × 10−3	4.49 × 10−3
5000	6.55 × 10−4	8.18 × 10−4	9.68 × 10−4	9.85 × 10−4	9.31 × 10−4
10000	3.28 × 10−4	4.13 × 10−4	4.89 × 10−4	4.95 × 10−4	4.75 × 10−4
15000	2.18 × 10−4	2.76 × 10−4	3.29 × 10−4	3.31 × 10−4	3.18 × 10−4
100000	3.28 × 10−5	4.21 × 10−5	-	4.99 × 10−5	4.98 × 10−5
500000	6.62 × 10−6	8.54 × 10−6	-	9.99 × 10−6	-