The population of the analytic layer

Builds, from a Weibull AFT DGM made by inject_mrct_structure(), the per-patient quantities the analytic layer works on: each super-population patient's log hazard under control ($\theta_0$) and log hazard ratio ($\psi$, so $\theta_1 = \theta_0 + \psi$), the H* (flag_harm) and region indicators, the Weibull mu and tau, and each patient's entry window: entry at $a_i + E$, $E \sim U(0, R)$, analysis at calendar time analysis_time.

Usage

ahr_population(dgm, analysis_time, max_entry, entry_delay = NULL)

# S3 method for class 'ahr_population'
print(x, ...)

Arguments

dgm: An aft_dgm_flex object made by inject_mrct_structure().
analysis_time: Calendar time of the analysis, in the DGM's time units.
max_entry: Length R of each patient's uniform entry window.
entry_delay: NULL (no delay), a non-negative numeric vector of length nrow(dgm$df_super) (each patient's delay $a_i$), or list(var = <column>, delay = a): delay a for patients with dgm$df_super[[var]] == 1, none otherwise (e.g. list(var = "z_region", delay = 6): the region enrols 6 months later). The analytic counterpart of simulate_from_dgm(entry_delay = ).
x: An ahr_population.
...: Unused.

Value

An object of class ahr_population: a list with theta0, psi, band (H*), region, delay, mu, tau, x (analysis_time), R (max_entry), N, the distinct delays, and an environment holding the quadrature grids and Laplace-sum caches (keyed by horizon).

Details

The DGM's theta_0, theta_1 and loghr_po columns are checked to be consistent with its linear predictors and with each other (as the mrct-consistency ahr_pop() did); the layer reads the hazards from them and uses dgm$model_params$mu as it stands, so apply scale_dgm_hazard() before building the population.

Limits

Weibull AFT DGMs from inject_mrct_structure() only; entry uniform on $(a_i, a_i + R)$ for each entry group; administrative censoring only (the random censoring model of the DGM, e.g. cens_adjust, is not in the layer); correlations exact for nested sets. Each distinct delay is one quadrature group with its own breaks, so a delay vector with many distinct values is slow. Derivations: the technical report (docs/manuscript/techreport/analytic-layer.qmd, Chapter 6, of the mrct-consistency repository) and the theory notes (docs/theory/AHR_mrct_dgm.qmd, docs/theory/AHR_complete.qmd).

Examples

if (FALSE) { # \dontrun{
pop <- ahr_population(dgm, analysis_time = 40, max_entry = 11)
pop6 <- ahr_population(dgm, 40, 11, entry_delay = list(var = "z_region", delay = 6))
} # }