breg {bayesm} | R Documentation |
breg
makes one draw from the posterior of a univariate regression
(scalar dependent variable) given the error variance = 1.0.
A natural conjugate (normal) prior is used.
breg(y, X, betabar, A)
y |
n x 1 vector of values of dep variable |
X |
n x k design matrix |
betabar |
k x 1 vector for the prior mean of the regression coefficients |
A |
k x k prior precision matrix |
model: y = X'β + e with e ~ N(0,1).
prior: β ~ N(betabar, A^{-1}).
k x 1 vector containing a draw from the posterior distribution
This routine is a utility routine that does not check theinput arguments for proper dimensions and type. In particular, X
must be a matrix. If you have a vector for X
, coerce itinto a matrix with one column.
Peter Rossi, Anderson School, UCLA, perossichi@gmail.com.
For further discussion, see Bayesian Statistics and Marketing by Rossi, Allenby, and McCulloch.
http://www.perossi.org/home/bsm-1
if(nchar(Sys.getenv("LONG_TEST")) != 0) {R=1000} else {R=10} ## simulate data set.seed(66) n = 100 X = cbind(rep(1,n), runif(n)); beta = c(1,2) y = X %*% beta + rnorm(n) ## set prior betabar = c(0,0) A = diag(c(0.05, 0.05)) ## make draws from posterior betadraw = matrix(double(R*2), ncol = 2) for (rep in 1:R) {betadraw[rep,] = breg(y,X,betabar,A)} ## summarize draws mat = apply(betadraw, 2, quantile, probs=c(0.01, 0.05, 0.50, 0.95, 0.99)) mat = rbind(beta,mat); rownames(mat)[1] = "beta" print(mat)