# ugarchroll arguments

The ugarchroll function is instrumental for rolling volatility prediction and avoiding look-ahead bias by conditioning the estimation to using only the returns available at the past time of estimation.

The ugarchroll gives the user flexibility in implementing the rolling estimation through the arguments n.start, refit.window and refit.every.

garchroll <- ugarchroll(tgarchspec, data = EURUSDret,
n.start = ___, refit.window = ___, refit.every = ___)


Which values should these arguments have in case the estimation sample consists of the 2000 most recent observations and the modeler only wants to re-estimate the model every 500 observations?