Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Stationarity and special considerations in dollar-bar vs time-bars

In considering stationarity and its importance to ML and building algos such as pair trades, much of the discussion and examples use time series data.

Such as Max's lecture: Basic Statistical Arbitrage: Understanding the Math Behind Pairs Trading
https://www.youtube.com/watch?v=g-qvFjvyqcs

Is there anything that should be considered, if rather than using time bars we use dollar bars? Can we still apply the same stationarity tests/transformations on dollar bar data? Can we still compare two different assets for arbitrage opportunities using dollar bars? If so, what needs to be modified?

I'm just starting to wade into this idea, so any articles or thoughts, would be appreciated.

1 response

@Mercer,
Looking at the dePrado book, there are some chapters addressing dollar bars, and some work was done here in the thread:
https://www.quantopian.com/posts/data-structures-for-financial-machine-learning
Going through that thread, with it's notebooks is very instructive!
alan