Dirichlet-Tree Allocation Priors for Bayesian Causal Forests

Key points

• The prior on the leaf values of the treatment trees determines whether sampling variability is attributed to the treatment effect.1
• We propose the Dirichlet-Tree Bayesian Causal Forest, which jointly regularizes the leaf values of each treatment tree through a Dirichlet-Tree prior on their variance shares, generated by stick-breaking along the tree topology.1
• We characterize the spike and tails of the marginal leaf prior, show that the prior adaptively bounds the number of regions of effect heterogeneity, and establish posterior consistency for the forest.1
• In simulations, the model outperforms BCF when heterogeneity is confined to a few regions and reduces spurious heterogeneity elsewhere, and, unlike independent shrinkage priors, its accuracy does not deteriorate as the trees grow.1

Sources (1)

• [1]Dirichlet-Tree Allocation Priors for Bayesian Causal Forests
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 04:57 PM

  Bayesian Causal Forests (BCF) estimate heterogeneous treatment effects with a separate ensemble of treatment trees.

  The prior on the leaf values of the treatment trees determines whether sampling variability is attributed to the treatment effect.

Extractive summary: sentences quoted from the sources.