Fitted BIc component parts under a counterfactual treatment value

Evaluates the fitted inflation probability/probabilities and continuous mean of a bic_fit() object with the treatment variable set to a fixed counterfactual value a for every row of newdata, using the fitted coefficients and the model's own design matrices directly (not predict.gamlss(..., newdata=), which re-resolves its stored data argument by name in the calling frame and silently returns the wrong frame when the fit was produced inside a function; see Ospina (2026), Section 4).

Usage

bic_parts(fit, treat, newdata, a)

Arguments

fit

A fitted model from bic_fit().

treat

Character, the name of the treatment variable in newdata.

newdata

A data frame with treat and all other covariates used by fit.

a

The counterfactual value to substitute for treat (0 or 1).

Value

A list with mu (continuous mean), p0 (probability of the lower boundary) and, for two-boundary fits, p1 (probability of the upper boundary), and EY, the mixture mean \(p_1 + (1 - p_0 - p_1)\mu\).