Parametric bootstrap and non-parametric permutations of a response vector or matrix

Create a random parametric bootstrap sample or a permutation of the input response vector or matrix (for survival outcomes).

SampleResponses(
  response_vec,
  event_vec = NULL,
  respType = c("survival", "regression", "categorical"),
  parametric = TRUE
)

SampleSurv(response_vec, event_vec, parametric = TRUE)

SampleReg(response_vec, parametric = TRUE)

SampleCateg(response_vec, parametric = TRUE)

Arguments

response_vec

The dependent vector to sample from. For survival response, this is the vector of event times. For regression or n-ary classification, this is the vector of responses.
event_vec

The death / event observation indicator vector for survival response. This is coded as 0 for a right-censoring occurence and 1 for a recorded event.
respType

What type of response has been supplied. Options are "none", "survival", "regression", and "categorical". Defaults to "none" to match the default response = NULL value.
parametric

Should the random sample be taken using a parametric bootstrap sample? Defaults to TRUE.

Value

If parametric = FALSE, a permutation of the supplied response is returned (for AES-PCA). If parametric = TRUE, we return a parametric bootstrap sample of the response.

Details

The distributions (for parametric = TRUE) are Weibull for survival times, Normal for regression response, and n-ary Multinomial for categorical response. Distributional parameters are estimated with their maximum likelihood estimates. When parametric = FALSE, the response vector or survival matrix is randomly ordered by row. This option should only be used when called from the AES-PCA method.

Examples

  # DO NOT CALL THESE FUNCTIONS DIRECTLY.
  # Use AESPCA_pVals() or SuperPCA_pVals() instead

if (FALSE) {
  data("colon_pathwayCollection")
  data("colonSurv_df")

  SampleResponses(
    response_vec = colonSurv_df$OS_time,
    event_vec = colonSurv_df$OS_event,
    respType = "survival"
  )
}