The fit with newReturns appended: the variance recursion runs on over
them at the fit's parameters, unchanged. The answer is a GarchFit —
forecastVariance, simulate and tailRisk take it as they take any fit —
whose variance path, lastResidual, log-likelihood, AIC and BIC cover the
whole series (k as estimated, n the whole length), and whose extendedBy
counts the returns added since the estimate. The parameters, inference,
modes and converged stay the estimate's.
For garch and gjr the variance path and log-likelihood are those of
logLikelihood over the concatenated series at the same parameters, to
the bit. EGARCH restarts from ln of the last variance, which can differ from
the log-variance the recursion held by one rounding: the same to ~1e-15.
The fit with
newReturnsappended: the variance recursion runs on over them at the fit's parameters, unchanged. The answer is aGarchFit—forecastVariance,simulateandtailRisktake it as they take any fit — whose variance path,lastResidual, log-likelihood, AIC and BIC cover the whole series (k as estimated, n the whole length), and whoseextendedBycounts the returns added since the estimate. The parameters,inference,modesandconvergedstay the estimate's.For garch and gjr the variance path and log-likelihood are those of
logLikelihoodover the concatenated series at the same parameters, to the bit. EGARCH restarts from ln of the last variance, which can differ from the log-variance the recursion held by one rounding: the same to ~1e-15.