predict_all: Obtain all possible predictions with one call
In osofr/growthcurveSL: SuperLearner for Imputing Growth Curves

This function creates a data.frame with one row per subject and a column named "fits". The latter column contains the 3 possible subject-level predictions: fit - training data model predictions, fitgrid - grid-based model predictions, holdout - predictions for holdouts or out-of-sample predictions. Some of these predictions might be empty, e.g. when add_holdout = FALSE, no holdout predictions are made.

predict_all(modelfit, newdata, add_holdout = TRUE, add_grid = TRUE, tgrid,
  grid_size = 150, add_checkpoint = TRUE, checkpoint = as.integer(c(1,
  hbgd::months2days(1:24))), add_MSE = TRUE,
  verbose = getOption("growthcurveSL.verbose"))

`modelfit`	Model fit object returned by `fit_growth`.
`newdata`	Subject-specific data for which predictions should be obtained.
`add_holdout`	Optional flag, set to `TRUE` to add holdout (out-of-sample) predictions used for scoring the current best model fit.
`add_grid`	Optional flag, set to `TRUE` to add a grid of equally spaced predictions over some range of the time variable for the current best model fit.
`tgrid`	Specify the grid of time-points directly. If missing a subject-specific grid is defined based on the subject's follow-up range.
`grid_size`	How many time-points should be used in equally spaced grid? This argument is only used when `tgrid` is missing.
`add_checkpoint`	Set to `TRUE` (default) to obtain predictions for pre-specified x checkpoint (essentially the same thing as grid, but with potentially different spacings).
`checkpoint`	The grid of checkpoint for which to obtain predictions (daily resolution).
`add_MSE`	Add subject-specific MSE for growth curve prediction (NOT IMPLEMENTED).
`verbose`	Set to `TRUE` to print messages on status and information to the console.

A data.frame with one row per subject. The column named "fit" contains nested subject-specific observed data and predictions. Each entry (cell) of "fit" is a list of 4 subject-specific data.frames, named "xy", "fit", "fitgrid" and "holdout", containing the training time-outcome values, the model predictions for training data, the grid of equally spaced model prediction and the holdout / out-of-sample predictions, respectively.

osofr/growthcurveSL documentation built on May 24, 2019, 4:56 p.m.