Currently studying on financial risk management. I want to test different methods of VaR estimation. I want to model volatility using a GARCH(1,1) model. My question is what should the size of the sample be, given the fact that I use intraday data and need a 1-day forecast.
Secondly, when should I calibrate the GARCH(1,1) model. Should I re-estimate the equation daily or do static forecasts and re-estimate the equation in a preselected horizon (e.g 1 year)?