REL_fixed: Compute REL for a fixed estimate.
In hoangtt1989/BICEP: Bi-linear algorithm for composite empirical likelihood

Compute REL for a fixed estimate.

REL_fixed(Z_mat, q = NULL, outlier_loop_arg = optim_outlier_control(),
  wts_init = NULL, vary_penalty = c("RB", "RB2", "none"), reuse_delta = F,
  RB_mu = 10, RB_tau = 2, RB2_ksi = 2, outer_eps = 1e-08,
  outer_rel_eps = 1e-04, dual_step = 2, outer_maxit = 1000,
  wts_beta_rep = 1, outer_tol_type = c("primal_dual", "fval", "gamma",
  "primal"), mirror_arg = optim_mirror_control(line_search = T),
  verbose = F)

`Z_mat`	a matrix of the score equations with observations in rows and variables in columns.
`q`	an integer specifying the number of detected outliers. Default is `floor(.1 * n)`.
`outlier_loop_arg`	control arguments passed to the `delta` optimization; see `optim_outlier_control`.
`wts_init`	an initial guess for the weights vector. Default is `1/n`.
`vary_penalty`	a string specifying the type of penalty variation.
`reuse_delta`	use the last estimate of `delta` in the current outer loop iteration.
`RB_mu`	the `mu` parameter for the residual balancing scheme.
`RB_tau`	the `tau` parameter for the residual balancing scheme.
`RB2_ksi`	the `ksi` parameter for the `RB2` scheme.
`outer_eps`	absolute tolerance required for outer loop convergence.
`outer_rel_eps`	relative tolerance required for outer loop convergence.
`dual_step`	initial penalty parameter.
`outer_maxit`	maximum number of outer loop iterations.
`wts_beta_rep`	the number of repetitions of the block coordinate descent update within each outer loop iteration.
`outer_tol_type`	the type of tolerance check for the outer loop.
`mirror_arg`	control arguments passed to the mirror descent optimization; see `optim_mirror_control`.
`verbose`	a boolean to allow console output.