You can use trailing unweighted historical volatility, exponentially weighted volatility (Riskmetrics), or GARCH to predict 1-day volatility from daily returns. There are R packages such as rugarch for GARCH. Are there packages in R, Python, or something else to predict daily volatility from past n-minute, for example 5-minute, returns? Since there are 78 5-minute returns from 9:30 to 4:00, a plausible thing to do is compute the standard deviation (SD) of the last few hundred 5-minute returns and scale that SD to predict the daily return SD. But how do you incorporate the overnight returns, which are more volatile than the intraday 5-minute returns?
Submitted October 06, 2020 at 02:11PM by Beliavsky