fusedTree - Fused Partitioned Regression for Clinical and Omics Data
Fit (generalized) linear regression models in each leaf
node of a tree. The tree is constructed using clinical
variables only. The linear regression models are constructed
using (high-dimensional) omics variables only. The
leaf-node-specific regression models are estimated using the
penalized likelihood including a standard ridge (L2) penalty
and a fusion penalty that links the leaf-node-specific
regression models to one another. The intercepts of the leaf
nodes reflect the effects of the clinical variables and are
left unpenalized. The tree, fitted with the clinical variables
only, should be constructed outside of the package with the
'rpart' 'R' package. See Goedhart and others (2024)
<doi:10.48550/arXiv.2411.02396> for details on the method.
