hyperpar_mod: Find Scale Parameter for modular regression

In sdPrior: Scale-Dependent Hyperpriors in Structured Additive Distributional Regression

Description

Find Scale Parameter for modular regression

Usage

hyperpar_mod(Z, K1, K2, A, c = 0.1, alpha = 0.1, omegaseq, omegaprob,
  R = 10000, myseed = 123, thetaseq = NULL, type = "IG",
  lowrank = FALSE, k = 5, mc = FALSE, ncores = 1, truncate = 1)

Arguments

Z	rows from the tensor product design matrix
K1	precision matrix1
K2	precision matrix2
A	constraint matrix
c	threshold from eq. (8) in Klein & Kneib (2016)
alpha	probability parameter from eq. (8) in Klein & Kneib (2016)
omegaseq	sequence of weights for the anisotropy
omegaprob	prior probabilities for the weights
R	number of simulations
myseed	seed in case of simulation. default is 123.
thetaseq	possible sequence of thetas. default is NULL.
type	type of hyperprior for tau/tau^2; options: IG => IG(1,theta) for tau^2, SD => WE(0.5,theta) for tau^2, HN => HN(0,theta) for tau, U => U(0,theta) for tau, HC => HC(0,theta) for tau
lowrank	default is FALSE. If TRUE a low rank approximation is used for Z with k columns.
k	only used if lowrank=TRUE. specifies target rank of low rank approximation. Default is 5.
mc	default is FALSE. only works im thetaseq is supplied. can parallel across thetaseq.
ncores	default is 1. number of cores is mc=TRUE
truncate	default is 1. If < 1 the lowrank approximation is based on on cumsum(values)/sum(values).

Value

the optimal value for theta

Author(s)

Nadja Klein

References

Kneib, T., Klein, N., Lang, S. and Umlauf, N. (2017) Modular Regression - A Lego System for Building Structured Additive Distributional Regression Models with Tensor Product Interactions Working Paper.

sdPrior documentation built on May 2, 2019, 8:57 a.m.