时间序列Dickey-Fuller测试报错：ValueError: too many values to unpack (expected 2)

Hey there! Let's break down exactly what's going on with your Dickey-Fuller test error, and clear up those confusing dftest index values once and for all.

First: What does adfuller() actually return? #

I’m assuming you’re using the adfuller() function from statsmodels.tsa.stattools (the standard tool for this test). This function returns a 6-element tuple with specific values in order:

• dftest[0]: The test statistic (the core number from the DF test)
• dftest[1]: The p-value (the key metric to judge stationarity)
• dftest[2]: The number of lags used in the test (to account for autocorrelation)
• dftest[3]: The number of valid observations included in the test
• dftest[4]: A dictionary of critical values (for 1%, 5%, and 10% significance levels)
• dftest[5]: The maximum information criterion (like AIC, used to pick optimal lags)

Why you’re getting the too many values to unpack error #

Your error almost certainly comes from trying to unpack a subset of this tuple into fewer variables than elements exist. For example:
If your code looks like this:

stat, p_val = dftest[0:4]  # dftest[0:4] gives 4 values, but you're trying to put them into 2 variables

That’s a problem—4 values can’t fit into 2 variables, hence the ValueError.

How to fix it #

You have two clean ways to handle this:

Option 1: Unpack all values at once (most readable)

from statsmodels.tsa.stattools import adfuller

# Run the test on your cleaned, NaN-free data
dftest = adfuller(df['your_time_series_column'])

# Unpack all 6 values explicitly
test_stat, p_val, used_lags, n_obs, crit_vals, max_ic = dftest

Option 2: Index directly to get specific values

If you only need certain metrics, call them by their index without unpacking:

# Get just the test statistic and p-value
test_stat = dftest[0]
p_val = dftest[1]

# Get the critical values (e.g., 5% level)
critical_value_5pct = dftest[4]['5%']

Quick guide to interpreting each value #

• dftest[0] (test statistic): The larger the absolute value, the stronger the evidence against the "non-stationary" null hypothesis.
• dftest[1] (p-value): If this is less than 0.05 (your typical significance level), you can reject the null hypothesis and conclude your data is stationary.
• dftest[4] (critical values): If your test statistic is smaller than the critical value for your chosen significance level, you also reject the null hypothesis.

Full working example #

Here’s a complete snippet to adapt to your data:

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

# Load and clean your data (you already did the NaN removal step)
df = pd.read_csv('your_data.csv')
clean_df = df.dropna(subset=['your_time_series_column'])

# Run ADF test
dftest = adfuller(clean_df['your_time_series_column'])

# Print results clearly
print(f"ADF Test Statistic: {dftest[0]:.4f}")
print(f"P-Value: {dftest[1]:.4f}")
print(f"Lags Used: {dftest[2]}")
print(f"Critical Values: {dftest[4]}")

# Make a stationarity judgment
if dftest[1] < 0.05:
    print("\n✅ Data is stationary (reject null hypothesis)")
else:
    print("\n❌ Data is non-stationary (fail to reject null hypothesis)")

内容的提问来源于stack exchange，提问作者Josh