Multiply every hazard of a Weibull AFT DGM by a constant

Sets the AFT intercept of a DGM to mu_fitted - tau * log(m), where mu_fitted is the intercept as fitted when the DGM was built. Under the Weibull AFT model the baseline cumulative hazard is \(\Lambda_0(s) = (s e^{-\mu})^{1/\tau}\), so every patient's hazard under either arm is multiplied by m and every hazard ratio (individual, AHR, CDE) is unchanged; only event probabilities in a finite window move. The scaling is not cumulative: it is always applied to the fitted intercept, so scale_dgm_hazard(scale_dgm_hazard(dgm, 2), 3) equals scale_dgm_hazard(dgm, 3).

Usage

scale_dgm_hazard(dgm, m)

Arguments

dgm

An aft_dgm_flex object, typically from inject_mrct_structure().

m

Positive number: the hazard multiplier. m = 1 restores the fitted intercept (for an unscaled DGM, the DGM is returned unchanged except for the record).

Value

The DGM with model_params$mu set, the multiplier recorded in dgm$mrct$hazard_multiplier and the fitted intercept in dgm$mrct$mu_fitted, and the attribute "m" set to m (as the mrct-consistency helper ideal_set_m() sets it, so scripts reading attr(dgm, "m") keep working).

Details

The fitted intercept is taken, in order, from dgm$mrct$mu_fitted (stored by an earlier call), attr(dgm, "mu_fitted") (the attribute the mrct-consistency builders set), or, for a DGM never scaled, dgm$model_params$mu. A DGM that records a multiplier but no fitted intercept is refused.

See also

inject_mrct_structure(), ahr_population()

Examples

if (FALSE) { # \dontrun{
dgm3 <- scale_dgm_hazard(dgm, 3)   # every hazard x 3, HRs unchanged
} # }