macrocephalopod

#fintwit hivemind when ya'll calc correlations between crypto (or any 7-days-a-week asset) returns and SPY, how do you handle the weekends? Impute a zero for SPY? Or impute NA for crypto? Or something else? cc @GestaltU @alphaarchitect @choffstein @daniel_egan

So here's a thing that doesn't answer your question, but you can think of it as a jumping off point. Imagine you have two securities: A trades from t1 to t2, and B trades from t3 to t4. t3>t2 so the trading periods are disjoint.

Get daily returns for A: t2 -> t2' (where ' denotes next day) And daily returns for B: t3 -> t3'. Calculate corr.

This will be biased down because (assuming martingales) you're correlating A's t2 -> t3 return with B's "zero" (in expectation) return. Same for B's t2' -> t3 return with A's "zero" over the same time.