Optional ReadonlydistThe innovations: normal (default, arch's) or Student-t with ν fitted.
Optional ReadonlyexogenousGARCH-X (garch and gjr): x_t ≥ 0 aligned with the returns — an implied variance, a realised measure — entering s_{t+1} as δ·x_t, δ ≥ 0 fitted.
Optional ReadonlyinferenceAlso compute robust (Bollerslev–Wooldridge) standard errors and two-sided p-values.
Optional ReadonlymeanA fitted constant (default) or zero — μ held at 0.
Optional ReadonlymodelOptional ReadonlynuThe lowest ν a t fit may take (default T_NU_MIN); must exceed 2.
Optional ReadonlytargetVariance targeting (garch and gjr): ω is not estimated but tied to the sample variance, ω = var(r)·(1 − persistence), so the model's unconditional variance is the sample's. One parameter fewer.
Optional ReadonlywarmA previous fit to start from as well — the walk-forward hands each origin the last one.
fitGarch's specification: the model, its innovations and mean, and what else to compute.