calculate_m_eff: Compute inverse of 'solve_for_hiershrink_scale'

Description Usage Arguments Value

View source: R/calculate_m_eff.R

Description

Instead of providing a desired effective number of parameters, the user provides the scale value(s), which is c in the notation of Boonstra and Barbaro, and the the function gives the implied prior number of effective parameters based upon this. As with 'solve_for_hiershrink_scale', the user can provide one global scale parameter (scale1, leaving scale2 = NA) that applies to all parameters, or two regional scale parameters (scale1, scale2), that applies to a partition of the parameters as defined by the first npar1 parameters and the second npar2 parameters.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
calculate_m_eff(
  scale1,
  scale2 = NA,
  npar1,
  npar2 = 0,
  local_dof = 1,
  regional_dof = -Inf,
  global_dof = 1,
  slab_precision = (1/15)^2,
  n,
  sigma = 2,
  tol = .Machine$double.eps^0.5,
  max_iter = 100,
  n_sim = 2e+05
)

Arguments

scale1

global (if scale2=NA) or regional scale parameter value 1

scale2

regional scale parameter value 2, can be NA if scale 1 is specified

npar1

partition of the parameters

npar2

second part of partition of the parameters

local_dof

(pos. integer) numbers indicating the degrees of freedom for lambda_j and tau, respectively. Boonstra, et al. never considered local_dof != 1 or global_dof != 1.

regional_dof

regional degrees of freedom

global_dof

(pos. integer) numbers indicating the degrees of freedom for lambda_j and tau, respectively. Boonstra, et al. never considered local_dof != 1 or global_dof != 1.

slab_precision

(pos. real) the slab-part of the regularized horseshoe, this is equivalent to (1/d)^2 in the notation of Boonstra and Barbaro

n

sample size

sigma

varaince

tol

tolerance level

max_iter

max number of iterations

n_sim

number of simulates

Value

A list containing prior numbers 1 and 2, the implied prior number of effective parameters.


umich-biostatistics/AdaptiveBayesianUpdates documentation built on July 29, 2021, 3:06 a.m.