CausalForest.build_forestMethod

Build a causal forest.

  • if bootstrap=True we sample for each tree via bootstrap else we use subsampling
  • if honest=True we use 2 samples one too build splits and the other one to fill leaves otherwise we use the whole sample for the two steps
  • if const_mtry=True we use a constant mtry otherwise we use a random mtry following min(max(Poisson(m_pois),1),number_of_features)
  • if m_pois=-1 we set mtry to sqrt(number_of_features) else mtry is m_pois
  • if optimisatio=true we use cross validation to tune regression random forest of gamma
source
CausalForest.build_forest_olsMethod

Build a causal forest.

  • if centering=True Y and W are centered else they stay unchanged
  • if bootstrap=True we sample for each tree via bootstrap else we use subsampling
  • if honest=True we use 2 samples one too build splits and the other one to fill leaves otherwise we use the whole sample for the two steps
  • if const_mtry=True we use a constant mtry otherwise we use a random mtry following min(max(Poisson(m_pois),1),number_of_features)
source