How to Perform a Durbin-Watson Test in R

Spread the love

The Durbin-Watson (DW) test is a popular statistical test used to detect autocorrelation (serial correlation) in the residuals of a regression analysis. Autocorrelation can violate the assumptions of independent errors, which can lead to unreliable coefficient estimates and misleading significance tests.

In this article, we will explore the Durbin-Watson test, its significance, and a step-by-step guide on how to perform the test in R.

What is Autocorrelation?

Autocorrelation refers to the correlation between members of a series of numbers ordered in time or space. In the context of regression analysis, if the error terms (residuals) from the model are correlated, they display autocorrelation. This violates the assumption of independent errors, which is vital for valid hypothesis testing in standard linear regression models.

The Durbin-Watson Test

The DW test statistic is calculated as:


  • et​ is the residual at time tt
  • n is the number of observations

The DW statistic provides values between 0 and 4. A value of 2 suggests no autocorrelation. However, values deviating from 2 might indicate positive or negative autocorrelation:

  • DW < 2: Positive autocorrelation
  • DW > 2: Negative autocorrelation

Performing the Durbin-Watson Test in R

Step 1: Install and Load Necessary Packages

To perform the DW test in R, you will need the lmtest package. If you don’t have it installed, you can get it with:


Once installed, load the package:


Step 2: Create a Regression Model

For the purpose of this guide, we will use a simple linear regression model. Let’s assume you have a dataset data with dependent variable y and independent variable x.

model <- lm(y ~ x, data = data)

Step 3: Perform the Durbin-Watson Test

Now, using the dwtest() function from the lmtest package, you can perform the DW test on the regression model:

dw_result <- dwtest(model)

Step 4: Interpret the Results

The dwtest() function will provide a DW statistic and a p-value. The p-value will help you determine the statistical significance of the DW statistic:

  • p-value < 0.05: The test suggests autocorrelation in the residuals.
  • p-value > 0.05: The test doesn’t suggest significant autocorrelation in the residuals.

Moreover, inspect the DW statistic:

  • Close to 2: Likely no autocorrelation.
  • Significantly less than 2: Positive autocorrelation is likely.
  • Significantly more than 2: Negative autocorrelation is likely.


The Durbin-Watson test is a crucial tool for analysts and researchers using regression models. By identifying autocorrelation in residuals, one can better meet the assumptions of linear regression and thereby ensure the validity of the regression results.

If the Durbin-Watson test does suggest autocorrelation in your residuals, you might need to explore alternative modeling strategies or data transformations to address this issue. This could involve using time series models like ARIMA or the inclusion of lagged variables in your regression model.

Remember, always ensure your model meets its assumptions before drawing conclusions or making predictions!

Posted in RTagged

Leave a Reply