Predict biomarker trajectories

Usage

# S3 method for class 'epikinetics_fit'
predict(
  object,
  newdata = NULL,
  times = 0:150,
  type = c("population", "individual", "participant", "new"),
  participants = NULL,
  biomarkers = NULL,
  summary = TRUE,
  ndraws = NULL,
  chunk_size = 20L,
  max_rows = 5e+06,
  probs = c(0.025, 0.975),
  scale = c("response", "model"),
  include_observation_noise = FALSE,
  seed = NULL,
  ...
)

Arguments

object: An epikinetics_fit object.
newdata: Participant-level covariate profiles for population or new- participant predictions. NULL uses prediction_grid(): observed categorical combinations with continuous predictors fixed at their participant-level medians. The resulting predictions are conditional, not marginalised over the fitted covariate distribution.
times: Non-negative times since exposure.
type: "population" excludes participant variation; "individual" uses posterior effects for fitted participants; and "new" draws new participant effects for each posterior draw/profile. "participant" is retained as an alias for "individual".
participants: Optional fitted participant ids for individual prediction. NULL selects all fitted participants.
biomarkers: Optional biomarker subset. The stored biomarker order is retained.
summary: Return posterior summaries rather than individual draws.
ndraws: Optional maximum number of posterior draws.
chunk_size: Number of fitted participants processed together. Summary predictions are calculated in bounded chunks and do not materialise the complete participant-by-draw-by-time table.
max_rows: Safety limit for unsummarised output. Subset participants, biomarkers, times, or posterior draws to stay below this value; use Inf only when the resulting memory requirement has been considered explicitly.
probs: Lower and upper interval probabilities.
scale: Return values on the natural response or model log2 scale.
include_observation_noise: Include residual measurement error.
seed: Optional seed for new-participant effects or observation noise.
...: Reserved for future methods.

Value

An epikinetics_prediction data frame. Unsummarised predictions include .draw; summaries include estimate (an alias for median), mean, median, lower, and upper. Without observation noise, intervals describe uncertainty in the latent expected trajectory. With observation noise, they are posterior predictive intervals for a future measurement.