opcg_made: OPCG-MADE - Local gradient estimation
In HarrisQ/linearsdr: Linear Sufficient Reduction Methods

View source: R/opcg_wrap_cpp.R

opcg_made

R Documentation

OPCG-MADE - Local gradient estimation

Description

This is an internal function called by OPCG. MADE also uses this function in its OPCG-step. This estimates the local intercept and slope coefficients.

Usage

opcg_made(
  x_matrix,
  y_matrix,
  bw,
  lambda,
  B_mat = NULL,
  ytype = "continuous",
  method = "newton",
  parallelize = F,
  r_mat = NULL,
  control_list = list()
)

Arguments

`x_matrix`	a 'nxp' matrix of predictors;
`y_matrix`	a 'nxm' response;
`bw`	the bandwidth parameter for the kernel; the default kernel is gaussian
`lambda`	an L2 penalty term for the negative log-likelihood
`B_mat`	the fixed coefficient matrix in MADE-step of MADE; not needed for OPCG, i.e. is set to the identity
`ytype`	the response type; continuous, categorical or ordinal
`method`	"newton" or "cg" methods; for carrying out the optimization using the standard newton-raphson (i.e. Fisher Scoring) or using Congugate Gradients
`parallelize`	Default is False; to run in parallel, you will need to have foreach and some parallel backend loaded; parallelization is strongly recommended and encouraged.
`r_mat`	a 'pxd' matrix for refining the weights in rOPCG and rMADE
`control_list`	a list of control parameters for the Newton-Raphson or Conjugate Gradient methods

Value

ahat - List of estimated local intercepts
Dhat - List of estimated local slopes/gradients
Dhat_ls - List of initial values for local slopes/gradients; for least squares, these are the same as the Dhat
weights - The kernel weights used in the local-linear estimation;