Course contentsShow
← Data Science
Lesson 717 of 2,14519. Time Series FoundationsFree lesson

KPSS Test

Use the KPSS test as a complementary approach where the null hypothesis is stationarity rather than non-stationarity.

KPSS Test

You'll learn how the KPSS test flips the stationarity testing logic by assuming your time series is stationary until proven otherwise.

What Is the KPSS Test?

The Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test is a statistical test for stationarity that works opposite to the Augmented Dickey-Fuller test you just learned. While ADF assumes non-stationarity as the null hypothesis, KPSS assumes stationarity is the default.

Think of it like two detectives investigating the same case from different angles: ADF asks "Is this series non-stationary?" while KPSS asks "Is this series stationary?" Using both gives you a more complete picture.

Why Use KPSS Alongside ADF?

Because the tests have opposite null hypotheses, they complement each other beautifully:

  • KPSS rejects (low p-value) → Evidence against stationarity
  • KPSS fails to reject (high p-value) → Evidence for stationarity

When you combine both tests:

  • ADF rejects + KPSS doesn't reject = Strong evidence of stationarity ✓
  • ADF doesn't reject + KPSS rejects = Strong evidence of non-stationarity ✓
  • Both reject or both don't reject = Inconclusive, investigate further

Interpreting KPSS Results

The KPSS test statistic measures how far your data deviates from stationarity. A large test statistic (and therefore small p-value) means you reject the null hypothesis of stationarity—your series is likely non-stationary.

This reversed logic is powerful because it protects against false positives for stationarity. Just because ADF says "not non-stationary" doesn't guarantee stationarity; KPSS confirms it from the other side.

Key Takeaway: The KPSS test assumes stationarity as its null hypothesis, making it the perfect complement to ADF—use both tests together for robust stationarity assessment.