~hackernoon | Bookmarks (2)

I focus on methods to handle missing data in financial time series. Using some some example data I show that LOCF is usually a decent go-to method compared to dropping and imputation but has its faults - i.e. can create artificial undesirable jumps in data. However, alternatives like interpolation have their own problems especially in context of live prediction/forecasting.

The author experienced missing items in Bitwarden, which were actually items not assigned to any collection due to a recent system change. They provide a solution to fix this issue by assigning items to collections. Lessons learned include the importance of observability in system changes, understanding user behavior variations, preserving system states during updates, ensuring consistent behavior across interfaces, providing efficient tooling for users,...