Calculate State Occupancy Probabilities — soprob

Computes patient-level state occupancy probabilities from a supported Markov transition model. Frequentist first-order models use the optimized native engine; other supported workflows use the reference engine.

Usage

soprob_markov(
  model,
  newdata,
  times = NULL,
  y_levels,
  absorb = NULL,
  time_var = "time",
  p_var = "yprev",
  p2_var = NULL,
  gap_var = NULL,
  time_covariates = NULL,
  include_re = FALSE,
  id_var = NULL,
  n_draws = 100L,
  seed = NULL,
  ...
)

Arguments

model: A Markov transition model fitted by vglm_markov(), orm_markov(), or blrm_markov(). The robust robcov_vglm result returned by vglm_markov() is supported; raw backend fits without fitting- wrapper provenance are not.
newdata: A data frame containing one baseline prediction row per patient or profile.
times: Required consecutive visits, in increasing order. The recursion takes one transition per value, so leaving out visits does not skip time; see sops(). Unlike sops(), this function does not warn when times leaves out fitted visits.
y_levels: Ordered outcome-state levels. The number of levels must equal the fitted threshold count plus one.
absorb: Optional absorbing state levels.
time_var: Name of the model time variable.
p_var: Name of the previous-state variable.
p2_var: Optional second previous-state variable. A non-NULL value selects second-order recursion.
gap_var: Optional elapsed-time-gap variable.
time_covariates: Optional data frame with one row per requested visit containing precomputed time-varying covariates.
include_re: For blrm fits, include fitted random-effect draws for known IDs.
id_var: ID column used for blrm random effects.
n_draws: Number of blrm posterior draws, or NULL for all draws.
seed: Optional seed used when selecting posterior draws.
...: Reserved for internal posterior draw batching.

Value

An array of state probabilities. Frequentist fits return [patient, time, state]; Bayesian fits return [draw, patient, time, state].