In this section, we show how to implement the model of Bialkowski, J., Darolles, S., and Le Fol, G. (2008) in R. We cover every detail, from loading the data to estimating the model parameters and producing the actual forecasts.
The data we use consists of 10 different stocks from the Dow Jones Industrial Average index (see the next table for an overview). We use the 21 trading days between 06/01/2011 and 06/29/2011. Trading on NYSE and NASDAQ is continuous between 09:30 and 16:00. After aggregating the data into 15-minute time slots, we receive 26 observations every day, and a total of 26 * 21 = 546 observations overall.
Tip
We divided the trading day into 26 time slots, whereas the original article defined 25. This is due to the difference in the opening hours of the different markets from where data was drawn. This only changes one single parameter in the model, but some attention must be paid to this detail.
All the used stocks are liquid enough to have positive...