Generate Augmented Repeated Measures Data for Pooled Hazards Regression

Generate Augmented Repeated Measures Data for Pooled Hazards Regression

format_long_hazards(
  A,
  W,
  wts = rep(1, length(A)),
  grid_type = c("equal_range", "equal_mass"),
  n_bins = NULL,
  breaks = NULL
)

Arguments

A

The numeric vector or similar of the observed values of an intervention for a group of observational units of interest.

W

A data.frame, matrix, or similar giving the values of baseline covariates (potential confounders) for the observed units whose observed intervention values are provided in the previous argument.

wts

A numeric vector of observation-level weights. The default is to weight all observations equally.

grid_type

A character indicating the strategy (or strategies) to be used in creating bins along the observed support of the intervention A. For bins of equal range, use "equal_range"; consult documentation of cut_interval for more information. To ensure each bin has the same number of points, use "equal_mass"; consult documentation of cut_number for details.

n_bins

Only used if grid_type is set to "equal_range" or "equal_mass". This numeric value indicates the number(s) of bins into which the support of A is to be divided.

breaks

A numeric vector of break points to be used in dividing up the support of A. This is passed through the ... argument to cut.default by cut_interval or cut_number.

Value

A list containing the break points used in dividing the support of A into discrete bins, the length of each bin, and the reformatted data. The reformatted data is a data.table of repeated measures data, with an indicator for which bin an observation fails in, the bin ID, observation ID, values of W for each given observation, and observation-level weights.

Details

Generates an augmented (long format, or repeated measures) dataset that includes multiple records for each observation, a single record for each discretized bin up to and including the bin in which a given observed value of A falls. Such bins are derived from selecting break points over the support of A. This repeated measures dataset is suitable for estimating the hazard of failing in a particular bin over A using a highly adaptive lasso (or other) classification model.