Problem
Form E-mini S&P 500 futures dollar bars:
(a) Compute Bollinger bands of width 5% around a rolling moving average. Count how many times prices cross the bands out (from within the bands to outside the bands).
(b) Now sample those bars using a CUSUM filter, where {yt} are returns and h = 0.05. How many samples do you get?
(c) Compute the rolling standard deviation of the two-sampled series. Which one is least heteroscedastic? What is the reason for these results?