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.