estimate_model: Estimate Model

In itamarfaran/corrfuncs: Analysis of populations of correlation matrices

Description

todo: explain

Usage

estimate_model(
  control_arr,
  diagnosed_arr,
  dim_alpha = 1,
  LinkFunc = LinkFunctions$multiplicative_identity,
  model_reg_config = list(),
  matrix_reg_config = list(),
  iid_config = list(),
  cov_config = list(),
  raw_start = FALSE,
  bias_correction = FALSE,
  early_stop = FALSE,
  verbose = TRUE
)

Arguments

control_arr	array of control group correlation matrices. either an array or data matrix form
diagnosed_arr	array of diagnosed group correlation matrices. either an array or data matrix form
dim_alpha	the number of columns in alpha. default 1
LinkFunc	a list of function. must include func, inverse, rev_func and null_value. see LinkFuncSkeleton
model_reg_config	see configurations. arguments passed will override the defaults.
matrix_reg_config	see configurations. arguments passed will override the defaults.
iid_config	list of two lists named 'iter_config' and 'optim_config', for the optimization of the model with identity matrix covariance matrix. see configurations. arguments passed will override the defaults.
cov_config	list of two lists named 'iter_config' and 'optim_config', for the optimization of the model with a specified covariance matrix. see configurations. arguments passed will override the defaults.
raw_start	if true, don't optimize with the identity matrix before optimizing with a specified covariance matrix
bias_correction	if true, correct the estimates to the median: a' = a - med(a) + null_value
early_stop	if true, stop the optimization of the joint loss function (of theta and alpha) didn't decrease in the last iteration.
verbose	if true, print status to console

Value

a list of the following:

• theta: a matrix, the estimates of theta

• alpha: a matrix, the estimates of alpha

• LinkFunc: the link function used same as the parameter LinkFunc

• regularization: the same as model_reg_config parameter

• vcov: the weighting matrix used in the optimization - the inverse of the correlations' covariance matrix

• convergence: a vector with the convergence in each iteration. see optim

• steps: the estimates of theta, alpha in each iteration

• log_optim: if optim_config$log_optim=TRUE, will return the output of optim for each iteration. else, NA.

See Also

configurations

itamarfaran/corrfuncs documentation built on Nov. 26, 2021, 12:02 p.m.