ADF Test in Time Series Python: What It Is and How to Use

The ADF test (Augmented Dickey-Fuller test) in Python is a statistical test used to check if a time series is stationary, meaning its properties do not change over time. It helps decide if differencing or other transformations are needed before modeling time series data.

How It Works

The ADF test checks if a time series has a unit root, which means it shows trends or patterns that change over time. Imagine watching the temperature every day: if it keeps rising or falling, it is not stationary. The ADF test tries to find out if such trends exist.

It does this by fitting a simple model that predicts the current value based on past values and checks if the series tends to return to a constant mean or drifts away. If it returns, the series is stationary; if it drifts, it is not.

Think of it like checking if a ball in a bowl settles at the bottom (stationary) or keeps rolling away (non-stationary).

Example

This example shows how to run the ADF test on a sample time series using Python's statsmodels library and interpret the results.

python

import numpy as np
import pandas as pd
from statsmodels.tsa.stattools import adfuller

# Create a non-stationary time series (random walk)
np.random.seed(0)
random_walk = np.cumsum(np.random.normal(size=100))
time_series = pd.Series(random_walk)

# Run ADF test
result = adfuller(time_series)

print(f"ADF Statistic: {result[0]:.4f}")
print(f"p-value: {result[1]:.4f}")
print("Critical Values:")
for key, value in result[4].items():
    print(f"   {key}: {value:.4f}")

Output

ADF Statistic: -1.2884
p-value: 0.6273
Critical Values:
   1%: -3.4983
   5%: -2.8912
   10%: -2.5828

When to Use

Use the ADF test when you want to check if your time series data is stationary before applying models like ARIMA or other forecasting methods. Stationarity is important because many models assume the data's behavior does not change over time.

For example, if you analyze sales data that grows over time, the ADF test can tell you if you need to remove trends or seasonality first. It is commonly used in finance, weather forecasting, and any field involving time-based data.

Key Points

• The ADF test checks if a time series is stationary by testing for a unit root.
• A low p-value (usually below 0.05) means the series is stationary.
• It helps decide if data needs transformation before modeling.
• Implemented in Python via statsmodels.tsa.stattools.adfuller.

Key Takeaways

The ADF test checks if a time series is stationary by detecting a unit root.

A p-value below 0.05 indicates the series is likely stationary.

Stationarity is crucial for many time series models to work well.

Use Python's statsmodels library to easily run the ADF test.

Transform non-stationary data before modeling to improve accuracy.