NormInvGamPosteriorSample: Generate conjugate posterior sample of coefficients for a...

In pjnewcombe/R2BGLiMS: R interface to the java package BGLiMS (Bayesian Generalised Linear Model Selection).

Description

This function can be used to generate a posterior sample of effects under the Normal-Inverse-Gamma conjugate model, for a particular combination of covariates (i.e. conditional on a fixed model). For use with JAM specify xTx and z instead of data and outcome.var. Note to self: adapted from Michael Jordan's lecture notes "Bayes Factors, g-priors, and Model Selection for Regression". Conugate expression is given under the 1/sigma prior. For the InverseGamma(a, b) prior as below, simply add a and b to the respective hyperparameters in equation (6) of Michael Jordan's notes.

Usage

NormInvGamPosteriorSample(
  data = NULL,
  outcome.var = NULL,
  confounders = NULL,
  model = NULL,
  tau = NULL,
  n.samples = 1000,
  xTx = NULL,
  z = NULL,
  sigma2_invGamma_a = NULL,
  sigma2_invGamma_b = NULL
)

Arguments

data	Matrix or dataframe containing the data to analyse. Rows are indiviuals, and columns contain the variables and outcome. If modelling summary statistics specify X.ref, marginal.betas, and n instead (see below).
outcome.var	Name of outcome variable in data. For survival data see times.var below. If modelling summary statistics with JAM this can be left null but you must specify X.ref, marginal.beats and n instead (see below).
confounders	Optional vector of confounders to fix in the model at all times, i.e. exclude from model selection.
model	Vector of covariate names to include in the model. Do not include confounders here - they should be specified with the confounders argument.
tau	Value to use for the g-prior sparsity parameter (tau*sigma^2 parameterisation).
n.samples	Number of posterior samples to draw.
sigma2_invGamma_a	Inverse-Gamma parameter a for the residual variance. Not specifying means the value in default.arguments is used (type "data(DefaultArguments)").
sigma2_invGamma_b	Inverse-Gamma parameter b for the residual variance. Not specifying means the value in default.arguments is used (type "data(DefaultArguments)").

Value

The posterior sample as a matrix. Rows are different posterior samples, and columns are parameters.

Author(s)

Paul Newcombe

pjnewcombe/R2BGLiMS documentation built on Feb. 10, 2020, 8:52 p.m.