Construct default covariate profiles for prediction

Builds inspectable participant-level profiles from the formula and model frame stored by prepare_epikinetics_data(). By default, categorical profiles are the combinations observed among fitted participants; this avoids silently extrapolating to unsupported combinations. Continuous covariates are held at their participant-level median. Use categorical = "cartesian" to request every combination of fitted factor levels, or pass an explicit newdata data frame to predict().

Usage

prediction_grid(
  x,
  categorical = c("observed", "cartesian"),
  continuous = c("median", "mean")
)

Arguments

x

An epikinetics_data or epikinetics_fit object.

categorical

Use combinations "observed" in the participant data or the full "cartesian" product of fitted categorical levels.

continuous

Hold continuous covariates at their participant-level "median" (default) or "mean".

Value

An ordinary data frame with one row per prediction profile. The .profile column is a stable row identifier; remaining columns are the original variables used by the model formula.

Details

These are conditional profiles, not averages over the fitted covariate distribution. Interactions and transformed terms are subsequently evaluated with the original terms object and contrasts when predict() constructs the numeric design matrix.

Examples

dat <- utils::read.csv(
  system.file("extdata", "delta.csv", package = "epikinetics")
)
prepared <- prepare_epikinetics_data(
  dat,
  formula = ~ infection_history,
  lower_limit = 5,
  upper_limit = 2560
)
prediction_grid(prepared)
#>   .profile                 infection_history
#> 1        1                   Infection naive
#> 2        2 Previously infected (Pre-Omicron)