Time Series
What breaks when observations are not independent: a regression that finds a relationship between two series that have nothing to do with each other, standard errors that are wrong by a known factor, and a validation split that reports a model more than five times better than it is.
Sign in to take quizzes, earn XP, and unlock stages as you reach 90% mastery.
Stationarity and the Spurious Regression
25 min · 100 XPTwo series with no connection whatever, regressed on each other and found significant 82.8% of the time, the property of the data that causes it, and the one-line transformation that restores an honest test.
Open lesson →Sign in to take the 3-question quiz.
Autocorrelation and the AR Model
30 min · 120 XPMeasuring how long a series remembers, the simplest model that produces such memory, and the exact factor by which dependence inflates the uncertainty of an average.
Open lesson →Sign in to take the 3-question quiz.
Forecasting and the Split That Lies
30 min · 120 XPWhy a shuffled validation split reports a model more than five times better than it is, the naive forecast that beats most methods on a random walk, and what an honest evaluation of a forecaster looks like.
Open lesson →Sign in to take the 3-question quiz.