inst/tinytest/errors-and-warnings.R

In ncvreg: Regularization Paths for SCAD and MCP Penalized Regression Models

# Single lambda warning
n <- 100
p <- 10
X <- matrix(rnorm(n*p), n, p)
y <- rnorm(n)
expect_warning(ncvreg(X, y, lambda=0.1))
y <- Surv(rexp(n), rbinom(n, 1, 0.5))
expect_warning(ncvsurv(X, y, lambda=0.1))

# mfdr
n <- 100
p <- 10
X <- matrix(rnorm(n*p), n, p)
y <- rnorm(n) > 0
fit <- lm(y~X)
expect_error(mfdr(fit))
fit <- ncvreg(X, y, family='binomial', returnX=FALSE)
expect_error(mfdr(fit))

# ncvreg
n <- 100
p <- 10
X <- matrix(rnorm(n*p), n, p)
y <- rnorm(n)
expect_error(ncvreg(rnorm, y), 'must be a matrix')
expect_error(ncvreg(matrix(LETTERS[1:10], 10, 1), 1:10), 'numeric matrix')
expect_error(ncvreg(X, 'a'), 'must be numeric')
expect_error(ncvreg(X, y, penalty='MCP', gamma=0.5), 'greater than 1')
expect_error(ncvreg(X, y, penalty='SCAD', gamma=1.5), 'greater than 2')
expect_error(ncvreg(X, y, nlambda=1), 'at least 2')
expect_error(ncvreg(X, y, alpha=0), 'greater than 0')
expect_error(ncvreg(X, c(NA, y[-1])), 'missing data')
expect_error(ncvreg(X, y, penalty.factor=1:floor(p/2)), 'penalty.factor and X do not match')
expect_error(ncvreg(X, y, family='binomial'), 'non-binary')

# ncvfit
n <- 100
p <- 10
X <- matrix(rnorm(n*p), n, p)
y <- rnorm(n)
expect_error(ncvfit(rnorm, y), 'must be a matrix')
expect_error(ncvfit(matrix(LETTERS[1:10], 10, 1), 1:10, lambda=0.1), 'numeric matrix')
expect_error(ncvfit(X, 'a'), 'must be numeric')
expect_error(ncvfit(X, y, penalty='MCP', gamma=0.5), 'greater than 1')
expect_error(ncvfit(X, y, penalty='SCAD', gamma=1.5), 'greater than 2')
expect_error(ncvfit(X, y, alpha=0), 'greater than 0')
expect_error(ncvfit(X, c(NA, y[-1])), 'missing data')
expect_error(ncvfit(X, y, penalty.factor=1:floor(p/2)), 'penalty.factor and X do not match')

# permres
n <- 100
p <- 10
X <- matrix(rnorm(n*p), n, p)
y <- rnorm(n)
fit <- ncvreg(X, y)
expect_error(permres(fit, lambda='a'), 'must be numeric')
fit_nox <- ncvreg(X, y, returnX=FALSE)
expect_error(permres(fit_nox, lambda=0.1), 'returnX')