init_params_lm_hs: Initialize the parameters for a linear regression

In drkowal/rSTAR: MCMC and EM algorithms for Simultaneous Transformation and Rounding (STAR) Models

Initialize the parameters for a linear regression

Description

Initialize the parameters for a linear regression model assuming a horseshoe prior for the (non-intercept) coefficients. The number of predictors p may exceed the number of observations n.

Usage

init_params_lm_hs(y, X)

Arguments

y	n x 1 vector of data
X	n x p matrix of predictors

Value

a named list params containing

1. mu n x 1 vector of conditional means (fitted values)

2. sigma the conditional standard deviation

3. coefficients a named list of parameters that determine mu

Note

The parameters in coefficients are:

• beta: the p x 1 vector of regression coefficients

• sigma_beta: the p x 1 vector of regression coefficient standard deviations (local scale parameters)

• xi_sigma_beta: the p x 1 vector of parameter-expansion variables for sigma_beta

• lambda_beta: the global scale parameter

• xi_lambda_beta: the parameter-expansion variable for lambda_beta components of beta

Examples

# Simulate data for illustration:
sim_dat = simulate_nb_lm(n = 100, p = 5)
y = sim_dat$y; X = sim_dat$X

# Initialize:
params = init_params_lm_hs(y = y, X = X)
names(params)
names(params$coefficients)

drkowal/rSTAR documentation built on July 5, 2023, 2:18 p.m.