Decompose a time series into trend / seasonal / noise (STL with a shuffled-null calibration so pseudo-cycles don't masquerade as seasonality) and test the classic calendar effects — day-of-week and turn-of-month — with HAC p-values plus an honest fixed-window TOM backtest vs buy & hold. Use when ask
Decompose a time series into trend / seasonal / noise (STL with a shuffled-null calibration so pseudo-cycles don't masquerade as seasonality) and test the classic calendar effects — day-of-week and turn-of-month — with HAC p-values plus an honest fixed-window TOM backtest vs buy & hold. Use when asked "is there a calendar edge", "day-of-week effect", "turn-of-month", "seasonality", or "trend vs no