Time-Series Foundations
A track of P7 · Time Series & Forecasting.
Order is information: trend, seasonality, autocorrelation and stationarity — reading a series before daring to forecast it.
Sales are up twelve per cent on last month. Before anyone celebrates, ask the only question that matters: is that growth, or is it December? A series can be rising and seasonal and noisy all at once, and a number that mixes the three together supports almost any story you want to tell with it.
A time series is a table where the row order carries the meaning — and the first skill is resisting the urge to model it before you can describe it. This track is the describing. You will split a wiggly line into the three stories it is telling at once (where it is heading, what repeats, what is left over), measure how strongly a series remembers its own past, and learn the one property — stationarity — that decides whether the classical machinery applies at all.
Everything runs on small seeded series generated by the same compute the labs use, so every number a panel shows is a number an exercise can re-derive. The autocorrelation module is the track's practice anchor: by its end you will have built the estimator by hand, matched it against the library value, and used it to find a hidden season — the exact move analysts use to read an unfamiliar series in the wild.