Recover an SNRinfo-shaped data.frame directly from a fitted model

Every model type fitDetFun() produces is fit against SNR and a detected/missed outcome, so both are still there to pull back out – this is what makes the separate SNRinfo argument optional.

Usage

extractSNRinfo(model)

Arguments

model

A fitted detFun object.

Value

A data.frame with columns SNR and Detected.

Details

glm/gam/scam are fit via a plain formula against a data.frame (Detected ~ SNR), so model.frame() already returns exactly those two columns – the same mechanism predictDetFunList() already relies on for its default SNR grid.

vglm's capture-recapture formula uses a multi-observer response matrix rather than a single Detected column, so it needs its own path: SNR comes from the @x design matrix and Detected from the @y response matrix, picking out whichObserver's column – the same slot predictDetFun.vglm() already uses for its own SNR range. Two list entries can share one underlying vglm fit and differ only in @extra$whichObserver (as in the CommonGround vignette, comparing observer 1 vs observer 2 on the same model) – this reads whichObserver fresh from each model object, so that distinction is preserved correctly.