BinaryGibbs_fit: Fitting Bayesian Probit Model.

In zovialpapai/PolyGibbs: Bayesian Analysis of Binary & Polychotomous Response Data

Description

BinaryGibbs_fit does Implementation of Probit Regression for Binary Responses via data augmentation and Gibbs sampling.

Usage

BinaryGibbs_fit(Train_X, Train_Y, nIter, prior, burn_in, prior_mean, prior_var)

Arguments

Train_X	n X p matrix of continuous covarites of training set.
Train_Y	n X 1 vector of responses of training set. (takes values 0 or 1).
nIter	An integer, No of iterations for the Gibbs Sampler.
prior	choice of prior (1: diffuse prior,2: proper conjugate prior).
burn_in	An integer, No of iterations neglected at begining of the chain in calculation of posterior mean.
prior_mean	a (p+1) X 1 vector specifying mean of the normal conjugate prior, if prior = 2.
prior_var	a (p+1) X (p+1) matrix specifying variance of the normal conjugate prior, if prior = 2.

Value

beta_matrix a nIter X (p+1) matrix of beta estimate chains.

estimates a (p+1) X 1 vector of beta estimates.

Train_Accuracy a nIter X 1 vector of accuracy

Examples

set.seed(250)
require(truncnorm)
require(MASS)
N <- 500
x1 <- seq(-1, 1, length.out = N)
x2 <- rnorm(N, 0, 1)
D <- 3
X <- matrix(c(rep(1, N), x1, x2), ncol = D)
true_theta <- c(- 0.5, 3.3, 2)
p <- pnorm(X %*% true_theta)
y <- rbinom(N, 1, p)
N1  <- sum(y)  # Number of successes
N0  <- N - N1  # Number of failures
#Spliting The Data in Train and Test in 80:20 ratio
Train_ID = sample(1:nrow(X), round(nrow(X) * 0.8), replace = FALSE) # Train Data IDS
Train_X = X[Train_ID, -1] # Train Data Covariates
Test_X = X[-Train_ID, -1] # Test Data Covarites
Train_Y = y[Train_ID] # Train Data Response
Test_Y = y[-Train_ID] # Test Data Response
nIter = 10000
burn_in = round(nIter * 0.5)
prior = 2
prior_mean = rep(1, 3)
prior_var = diag(10, 3)
BinaryGibbs_fit(Train_X, Train_Y, nIter, prior, burn_in, prior_mean, prior_var )

zovialpapai/PolyGibbs documentation built on Dec. 9, 2019, 6:52 a.m.