hcmm_hyperpar: Generate a list of hyperparameters

In MixedDataImpute: Missing Data Imputation for Continuous and Categorical Data using Nonparametric Bayesian Joint Models

Description

Generates a list of hyperparameters for use in hcmm_impute. Specifying only hcmmdat or q AND cx will generate default values (see citation).

Usage

hcmm_hyperpar(hcmmdat = NULL, q = ncol(hcmmdat$Y), cx = hcmmdat$cx,
  alpha_a = 0.5, alpha_b = 0.5, beta_x_a = 0.5, beta_x_b = 0.5,
  beta_y_a = 0.5, beta_y_b = 0.5, tau_a = 0.5, tau_b = 0.5, v = q + 1,
  w = q + 2, Sigma0 = diag(1, q)/v, gamma = 1/cx, sigma2_0beta = 10)

Arguments

hcmmdat	An hcmm_data object
q	The number of continuous variables
cx	A length p vector (where p is the number of categorical variables). cx[j] is the number of distinct values taken by X[,j]
alpha_a,alpha_b	Gamma prior on top-level concentration parameter, where E(alpha) = alpha_a/alpha_b
beta_x_a,beta_x_b	Gamma prior on X model concentration parameter
beta_y_a,beta_y_b	Gamma prior on Y model concentration parameter
tau_a,tau_b	Gamma prior on coefficient precision parameters
v,w	Degree of freedom parameters in the hierarchical inverse-Wishart/Wishart prior
Sigma0	Centering matrix in the hierarchical inverse-Wishart/Wishart prior
gamma	Parameter of the symmetric Dirichlet priors in the product multinomial kernel. (Should be a length p vector.)
sigma2_0beta	Variance of the prior on B0

Value

A list of hyperparameters

MixedDataImpute documentation built on May 1, 2019, 9:29 p.m.