multilevel_kernel_qp: Re-weight control sub-groups to treated sub-group with kernel...

In ebenmichael/balancer: Matrix constraints for covariate balance

Re-weight control sub-groups to treated sub-group with kernel imbalance

Description

Re-weight control sub-groups to treated sub-group with kernel imbalance

Usage

multilevel_kernel_qp(
  X,
  trt,
  Z,
  kernel = kernlab::vanilladot(),
  lambda = 0,
  lowlim = 0,
  uplim = 1,
  scale_sample_size = T,
  exact_global = T,
  verbose = TRUE,
  eps_abs = 1e-05,
  eps_rel = 1e-05,
  ...
)

Arguments

X	n x d matrix of covariates
trt	Vector of treatment assignments
Z	Vector of group indicators with J levels
kernel	Kernel to compute balance measure with, default is the inner product
lambda	Regularization hyper parameter, default 0
lowlim	Lower limit on weights, default 0
uplim	Upper limit on weights, default 1
scale_sample_size	Whether to scale the dispersion penalty by the sample size of each group, default T
exact_global	Whether to enforce exact balance for overall population
verbose	Whether to show messages, default T
eps_abs	Absolute error tolerance for solver
eps_rel	Relative error tolerance for solver
...	Extra arguments for osqp solver

Value

• weights Estimated weights as a length n vector

• imbalance Imbalance in covariates as a d X J matrix

• global_imbalanceOverall imbalance in covariates, as a length d vector

ebenmichael/balancer documentation built on Jan. 17, 2024, 2:56 p.m.