StochTree::GaussianMultivariateRegressionLeafModel Class Reference

Marginal likelihood and posterior computation for gaussian homoskedastic constant leaf outcome model. More...

#include <leaf_model.h>

Public Member Functions
	GaussianMultivariateRegressionLeafModel (Eigen::MatrixXd &Sigma_0)
	Construct a new GaussianMultivariateRegressionLeafModel object.
double	SplitLogMarginalLikelihood (GaussianMultivariateRegressionSuffStat &left_stat, GaussianMultivariateRegressionSuffStat &right_stat, double global_variance)
	Log marginal likelihood for a proposed split, evaluated only for observations that fall into the node being split.
double	NoSplitLogMarginalLikelihood (GaussianMultivariateRegressionSuffStat &suff_stat, double global_variance)
	Log marginal likelihood of a node, evaluated only for observations that fall into the node being split.
Eigen::VectorXd	PosteriorParameterMean (GaussianMultivariateRegressionSuffStat &suff_stat, double global_variance)
	Leaf node posterior mean.
Eigen::MatrixXd	PosteriorParameterVariance (GaussianMultivariateRegressionSuffStat &suff_stat, double global_variance)
	Leaf node posterior variance.
void	SampleLeafParameters (ForestDataset &dataset, ForestTracker &tracker, ColumnVector &residual, Tree *tree, int tree_num, double global_variance, std::mt19937 &gen)
	Draw new parameters for every leaf node in tree, using a Gibbs update that conditions on the data, every other tree in the forest, and all model parameters.

Detailed Description

Marginal likelihood and posterior computation for gaussian homoskedastic constant leaf outcome model.

Constructor & Destructor Documentation

GaussianMultivariateRegressionLeafModel()

StochTree::GaussianMultivariateRegressionLeafModel::GaussianMultivariateRegressionLeafModel ( Eigen::MatrixXd &  Sigma_0 )

inline

Construct a new GaussianMultivariateRegressionLeafModel object.

Parameters

Sigma_0

Prior covariance, must have the same number of rows and columns as dimensions of the basis vector for the multivariate regression problem

Member Function Documentation

SplitLogMarginalLikelihood()

double StochTree::GaussianMultivariateRegressionLeafModel::SplitLogMarginalLikelihood	(	GaussianMultivariateRegressionSuffStat &	left_stat,
		GaussianMultivariateRegressionSuffStat &	right_stat,
		double	global_variance
	)

Log marginal likelihood for a proposed split, evaluated only for observations that fall into the node being split.

Parameters

left_stat	Sufficient statistics of the left node formed by the proposed split
right_stat	Sufficient statistics of the right node formed by the proposed split
global_variance	Global error variance parameter

NoSplitLogMarginalLikelihood()

double StochTree::GaussianMultivariateRegressionLeafModel::NoSplitLogMarginalLikelihood	(	GaussianMultivariateRegressionSuffStat &	suff_stat,
		double	global_variance
	)

Log marginal likelihood of a node, evaluated only for observations that fall into the node being split.

Parameters

suff_stat	Sufficient statistics of the node being evaluated
global_variance	Global error variance parameter

PosteriorParameterMean()

Eigen::VectorXd StochTree::GaussianMultivariateRegressionLeafModel::PosteriorParameterMean	(	GaussianMultivariateRegressionSuffStat &	suff_stat,
		double	global_variance
	)

Leaf node posterior mean.

Parameters

suff_stat	Sufficient statistics of the node being evaluated
global_variance	Global error variance parameter

PosteriorParameterVariance()

Eigen::MatrixXd StochTree::GaussianMultivariateRegressionLeafModel::PosteriorParameterVariance	(	GaussianMultivariateRegressionSuffStat &	suff_stat,
		double	global_variance
	)

Leaf node posterior variance.

Parameters

suff_stat	Sufficient statistics of the node being evaluated
global_variance	Global error variance parameter

SampleLeafParameters()

void StochTree::GaussianMultivariateRegressionLeafModel::SampleLeafParameters	(	ForestDataset &	dataset,
		ForestTracker &	tracker,
		ColumnVector &	residual,
		Tree *	tree,
		int	tree_num,
		double	global_variance,
		std::mt19937 &	gen
	)

Draw new parameters for every leaf node in tree, using a Gibbs update that conditions on the data, every other tree in the forest, and all model parameters.

Parameters

dataset	Data object containining training data, including covariates, leaf regression bases, and case weights
tracker	Tracking data structures that speed up sampler operations, synchronized with active_forest tracking a forest's state
residual	Data object containing the "partial" residual net of all the model's other mean terms, aside from tree
tree	Tree to be updated
tree_num	Integer index of tree to be updated
global_variance	Value of the global error variance parameter
gen	C++ random number generator

---

The documentation for this class was generated from the following file:

• include/stochtree/leaf_model.h