Fit a Mixed-Effects Model for Integrated Brier Score Against Alpha

Fits a linear mixed-effects model (or fixed-effects regression fallback) of Integrated Brier Score (IBS) or `log10(IBS)` against `alpha`, accounting for grouping across resampling splits (`inner_resamples_splits`). Returns the fitted model, tidy coefficients, a formatted equation string, and model predictions along the alpha spectrum.

Usage

fit_ibs_mixed_model(data, log_ibs = TRUE)

Arguments

data

A data frame containing tuning results from [tune_alpha_by_split()]. Must contain columns `alpha`, `IBS`, and `inner_resamples_splits`.

log_ibs

Logical. When `TRUE` (default), models `log10(IBS)`. When `FALSE`, models raw `IBS`.

Value

A list with class `"cbe_tune_mixed_model"` containing:

model

The fitted `merMod` or `lm` model object

equation

A character string with the fitted equation (e.g. `"log10(IBS) = -0.8712 + 0.0450 * alpha"`)

coefficients

A tibble with `term`, `estimate`, and standard error

fitted

A tibble of predicted values across the alpha continuum

data

The deduplicated per-split fit data used for modeling

Details

This dynamic fitting eliminates hard-coded regression coefficients in reports.