test glmnet with random data

randomData.R

library(glmnet) 

# Binary data
n = 1000
p = 100
nzc = trunc(p/10)
for(i in 1:100) {
  cat(i, " ")
  if (i %% 10 == 0) cat("\n")
  x = matrix(rnorm(n * p), n, p)
  ly = rbinom(n = n, prob = .5, size = 1)
  cvfit = cv.glmnet(x, ly, family = "binomial")
  # plot(cvfit)
  if (sum(coef(cvfit, s = "lambda.min") != 0) ==0)
    cat("Got you, ", i, "\n") # note that se is very small around max(lambda)
}

# Survival data
data(CoxExample) # load x, y
nr <- 30; nc <- 100
y1 <- y[1:nr, ]
for(i in 1:100) {
  x1 <- matrix(rnorm(nr*nc), nr=nr)
  cvfit = cv.glmnet(x1, y1, family = "cox")
  # plot(cvfit)
  if (sum(coef(cvfit, s = "lambda.min") != 0) ==0)
  cat("Got you, ", i, "\n")
}

# Continuous response data
nr=20; nc=100
for(i in 1:100) {
  cat(i, " ")
  if (i %% 10 == 0) cat("\n")
  y <- as.matrix(rnorm(nr)) 
  X <- matrix(rnorm(nr*nc), nr=nr); colnames(X) <- paste0("x", 1:ncol(X))
  cvfit <- cv.glmnet(X, y)
  # plot(cvfit)
  if (sum(coef(cvfit, s = "lambda.min") != 0) ==0)
  cat("Got you, ", i, "\n")
}