sharp (version 1.4.6)

Resample: Resampling observations

Description

Generates a vector of resampled observation IDs.

Usage

Resample(data, family = NULL, tau = 0.5, resampling = "subsampling", ...)

Value

A vector of resampled IDs.

Arguments

data

vector or matrix of data. In regression, this should be the outcome data.

family

type of regression model. This argument is defined as in glmnet. Possible values include "gaussian" (linear regression), "binomial" (logistic regression), "multinomial" (multinomial regression), and "cox" (survival analysis).

tau

subsample size. Only used if resampling="subsampling" and cpss=FALSE.

resampling

resampling approach. Possible values are: "subsampling" for sampling without replacement of a proportion tau of the observations, or "bootstrap" for sampling with replacement generating a resampled dataset with as many observations as in the full sample. Alternatively, this argument can be a function to use for resampling. This function must use arguments named data and tau and return the IDs of observations to be included in the resampled dataset.

...

additional parameters passed to the function provided in resampling.

Details

With categorical outcomes (i.e. "family" argument is set to "binomial", "multinomial" or "cox"), the resampling is done such that the proportion of observations from each of the categories is representative of that of the full sample.

Examples

## Linear regression framework
# Data simulation
simul <- SimulateRegression()

# Subsampling
ids <- Resample(data = simul$ydata, family = "gaussian")
sum(duplicated(ids))

# Bootstrapping
ids <- Resample(data = simul$ydata, family = "gaussian", resampling = "bootstrap")
sum(duplicated(ids))

## Logistic regression framework
# Data simulation
simul <- SimulateRegression(family = "binomial")

# Subsampling
ids <- Resample(data = simul$ydata, family = "binomial")
sum(duplicated(ids))
prop.table(table(simul$ydata))
prop.table(table(simul$ydata[ids]))

# Data simulation for a binary confounder
conf <- ifelse(runif(n = 100) > 0.5, yes = 1, no = 0)

# User-defined resampling function
BalancedResampling <- function(data, tau, Z, ...) {
  s <- NULL
  for (z in unique(Z)) {
    s <- c(s, sample(which((data == "0") & (Z == z)), size = tau * sum((data == "0") & (Z == z))))
    s <- c(s, sample(which((data == "1") & (Z == z)), size = tau * sum((data == "1") & (Z == z))))
  }
  return(s)
}

# Resampling keeping proportions by Y and Z
ids <- Resample(data = simul$ydata, family = "binomial", resampling = BalancedResampling, Z = conf)
prop.table(table(simul$ydata, conf))
prop.table(table(simul$ydata[ids], conf[ids]))

# User-defined resampling for stability selection
stab <- VariableSelection(
  xdata = simul$xdata, ydata = simul$ydata, family = "binomial",
  resampling = BalancedResampling, Z = conf
)