SL.flexsurv: Accelerated failure time models
In wuziyueemory/twostageSL: Two-Stage Super Learner Prediction Function

Description Usage Arguments Value Examples

This function implements accelerated failure time regression models to estimate the conditional survival, which is back-transformed to estimate the conditional mean. It leverages the flexsurv package to fit accelerated failure time models and uses numerical integration to back-transform. The numerical integration step can be unstable and we have tried to build in checks to guard against this. In particular, we first try to integrate with upper limit = Inf, but if that fails move to 1e8, which sometimes is able to provide a sane answer when upper limit = Inf fails. The function keeps trying smaller and smaller values, but will not go smaller than 1e6. In that case it returns a random number between 0 and randomUpper (default is maxY).

SL.flexsurv(
  Y,
  X,
  newX,
  family,
  obsWeights,
  dist = "gengamma",
  randomUpper = max(Y),
  ...
)

`Y`	A numeric outcome variable
`X`	A `data.frame` of covariates constituting the training sample
`newX`	A `data.frame` with the same column names and format as `X` constituting the validation sample.
`family`	Gaussian only
`obsWeights`	Observation-level weights (not currently used)
`dist`	Distribution for accelerated failure time model (defaults to generalized Gamma)
`randomUpper`	If numeric integration fails, for the purposes of stability while fitting the `SuperLearner`, a random number is selected uniformly between 0 and `randomUpper` (defaults to `max(Y)`).

pred: Predicted outcomes based on predictors in newX
fit: A list with named entries object (the fitted flexsurv model object), randomUpper.

# load cost data
data(cost_data)
# fit AFT model
fit_flexsurv <- SL.flexsurv(Y = cost_data$totalcost, X = cost_data[, c("female", "race")],
                              newX = cost_data[, c("female", "race")])
# get back predictions
pred_flexsurv <- predict(fit_flexsurv$fit, newdata = cost_data[,c("female", "race")])