tests/test_linear.R

In antiphon/vbss: Spike and Slab generalized linear regression (GLM) using variational approximation

### test logistic fitting
library(devtools)
library(vbglmss)

N <- 10
P <- 5
P0 <- 2
K <- 2
set.seed(1234)
## covariates: SS
X <- matrix(rnorm((P0+P)*N, 0, 2), nrow=N)
s <- c(rep(1,P), rep(0, P0))
w <- s*rnorm((P+P0))
## covariates: gaussian
Z <- matrix(rnorm(K*N), nrow=N)
b <- rnorm(K)
## response:
#
y <- rnorm( N ,c(X%*%w) + c(Z%*%b), 1)
## observed covariates:
colnames(X) <- paste0("X.", 1:(P+P0))
##
res<- vbglmss(y=y, X=X, Z=Z, verbose=TRUE, prior=list(pi=0.01))

plot(1:(P0+P), res$w$pi, col=s+2)