I have a large data set consisting of daily stock returns over 30 years and 7000 stocks. The time series do not align with each other across the set (i.e. dropmissing(data) returns no rows). If I compute the covariance matrix in pandas/Python, I get a result that reflects pairwise covariances for overlapping periods. How do I do the same thing with a DataFrame or TimeArray without computing at the pairwise level. Please leave aside the issue of whether the pandas result is actually sensible. I am just trying to understand how to deal with lots of missing data in Julia without degrading performance by hacking the calculation manually.
2 posts - 2 participants
