context("Gaussian")

sigma =3
set.seed(1)
n <- 100; p <- 20
x <- matrix(rnorm(n * p), n, p)
beta <- matrix(c(rep(2, 5), rep(0, 15)), ncol = 1)
y <- x %*% beta + rnorm(n)*sigma
xtest=matrix(rnorm(n * p), n, p)
ytest <- xtest %*% beta + rnorm(n)*sigma


fit1 <- uniLasso(x, y)

pred1 <- predict(fit1,xtest[1:10,],s=1) #predict on test data

# Two-stage variation where we carve off a small dataset for computing the univariate coefs.

cset=1:20
info = uniInfo(x[cset,],y[cset])
fit2s <- uniLasso(x[-cset,], y[-cset], info = info)

cvfit1 <- cv.uniLasso(x, y)
predcv1 <- predict(cvfit1,xtest[1:10,], s="lambda.min")

cvfit2s <- cv.uniLasso(x[-cset,], y[-cset], info = info)

# cv.uniLasso with same positivity constraints, but starting `beta`
# from univariate fits on the same data. With loo=FALSE, does not tend to do as well,
# probably due to overfitting.

cvfitpa <- cv.uniLasso(x, y, loo = FALSE)

# cv.uniLasso with no constraints, but starting `beta` from univariate fits.
# This is a version of the adaptive lasso, which tends to overfit, and loses interpretability.


cvfita <- cv.uniLasso(x, y, loo = FALSE, lower.limits = -Inf)

fitu <- cv.uniReg(x, y)
fitua <- cv.uniReg(x, y, hard.zero = FALSE)
ci <- ci.uniReg(x, y, B=20)

objects = enlist(fit1,pred1,fit2s,cvfit1,predcv1,cvfit2s,cvfitpa,cvfita,fitu,fitua,ci)
###saveRDS(objects, "saved_results/test_Gaussian.RDS")

expected  <- readRDS("saved_results/test_Gaussian.RDS")
for (x in names(objects)) {
    cat(sprintf("Testing %s\n", x))
    expect_equal(objects[[x]], expected[[x]])
}