Topic · Statistics & Econometrics
← All topicsTime Series
40 articles · 5 checkpoints · 26 deeper reads · 9 reference notes
A standalone topic: it is on no roadmap, so read it on its own terms.
Every article, in reading order
plant a flag as you finish eachRead these first
Two fingerprint plots that tell you, before you fit anything, roughly how many lags of a time series' own past actually matter, the first step of building an ARIMA model instead of guessing at one.
The disciplined three-step loop, identify, fit, check the leftovers, for building an ARIMA model without either underfitting the structure or overfitting the noise.
How to turn a day's worth of high-frequency price ticks into a single, far more accurate volatility number than squaring one close-to-close return, and why simply using more and more ticks eventually makes the estimate worse, not better.
A way to re-view a time series not as a sequence over time but as a mixture of hidden cycles of different lengths, so a repeating pattern in the data shows up as a spike you can point to, instead of something you have to spot by eye.
The regression that lets two cointegrated series drift apart in the short run but always adds a pull back toward their long-run relationship, the engine behind statistical-arbitrage pairs trading.
Then the rest