estimate_model_jacknife: Jackknife Estimation of Model

In itamarfaran/corrpops: Analysis of populations of correlation matrices

Description

todo: explain

Usage

estimate_model_jacknife(
  control_arr,
  diagnosed_arr,
  dim_alpha = 1,
  alpha0 = NULL,
  theta0 = NULL,
  LinkFunc = LinkFunctions$multiplicative_identity,
  model_reg_config = list(),
  matrix_reg_config = list(),
  iid_config = list(iter_config = list(minit = 0)),
  cov_config = list(),
  return_gee = FALSE,
  jack_control = TRUE,
  bias_correction = FALSE,
  early_stop = FALSE,
  verbose = TRUE,
  ncores = 1
)

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
alpha0	starting point for alpha in the optimization. if null (the default), will use LinkFunc$null_value
theta0	starting point for alpha in the optimization. if null (the default), will the average matrix of all subjects
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.
return_gee	if true, calculate the gee estimate of variance in each jackknife
jack_control	if false, don't jackknife control subjects
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
ncores	number of cores to use in parallel

Value

a list of the following:

• theta: a matrix, the estimates of theta from each jackknife iteration vectorized

• alpha: a matrix, the estimates of alpha from each jackknife iteration vectorized

• gee_var: if return_gee=TRUE, an array of the gee estimate of variance from each jackknife iteration

• convergence: a matrix consisting of vectors with the convergence in each iteration. see optim

• is_diagnosed: a vector with 0,1 indicating of a control or diagnosed subjet was omitted in the iteration

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

• regularization: the same as model_reg_config parameter

See Also

estimate_model

itamarfaran/corrpops documentation built on Dec. 20, 2021, 8:02 p.m.