Table of Contents
Introduction
A Durbin-Watson result can look confusing when you do not know which critical value to use.
The Durbin Watson table helps you evaluate whether residuals show signs of autocorrelation in regression analysis. This matters when you are working with statistical results and need a reliable way to support your conclusion.
For beginners in the USA, the table can seem difficult at first. Terms like dL, dU, sample size, predictors, and significance level can make the process feel more complicated.
The good news is that using the table becomes much easier once you understand what each value means. You also need to know how your Durbin-Watson statistic fits between the critical values.
In this guide, you will learn what the Durbin-Watson test measures and how its table works. We will explain Durbin-Watson critical values in simple terms.
You will also learn how to find n and k, choose the correct significance level, and compare your statistic with dL and dU.
By the end, you will know how to use and read a Durbin-Watson table with confidence.
What Is the Durbin-Watson Test?
The Durbin-Watson test checks whether nearby residuals in a regression model are correlated. Residuals are the differences between your actual and predicted values.
This matters because regression models often assume that residuals are independent. When residuals influence each other, the model may have autocorrelation, also called serial correlation.
Autocorrelation means that one residual tends to relate to another residual nearby in the sequence. For example, a positive pattern may occur when several residuals stay above or below zero in a row.
The Durbin-Watson test uses a statistic ranging from 0 to 4. Its value gives you an initial idea about the direction of residual autocorrelation.
What Does the Durbin-Watson Statistic Tell You?
The Durbin-Watson statistic provides a quick indication of possible autocorrelation in regression residuals. Its value is commonly interpreted around the midpoint of 2.
- Around 2: Suggests little or no autocorrelation.
- Below 2: Suggests a tendency toward positive autocorrelation.
- Above 2: Suggests a tendency toward negative autocorrelation.
These ranges provide a useful starting point, but they do not give a complete hypothesis-test conclusion. A statistic slightly below or above 2 does not automatically prove autocorrelation.
For a formal decision, compare the statistic with the appropriate Durbin-Watson critical values. These values include the lower bound (dL) and upper bound (dU), based on factors such as sample size, predictors, and significance level.
This critical-value approach helps determine whether the evidence supports autocorrelation, no autocorrelation, or an inconclusive result.
What Is a Durbin-Watson Table?
A Durbin-Watson table lists critical values used to evaluate a Durbin-Watson test result. It helps you decide whether your regression residuals show evidence of autocorrelation.
The table uses two important boundaries: dL and dU. These values create different decision regions when you compare them with your calculated Durbin-Watson statistic.
- dL (lower bound): The lower critical value for the test.
- dU (upper bound): The upper critical value for the test.
- Significance level (α): The selected risk level for the statistical test, such as 0.05.
- Sample size (n): The number of observations included in your regression analysis.
- Number of predictors (k): The number of independent variables in your regression model.
To use the table correctly, you need to identify your n, k, and significance level first. You then find the matching dL and dU values.
The relationship between your Durbin-Watson statistic and these critical values guides the formal test decision. Some results may fall into an inconclusive region.
Understanding these table components makes the next step much easier. Once you know what each value represents, you can locate the correct row and column with confidence.
Durbin-Watson Critical Values Table
The Durbin-Watson critical values table helps you determine whether regression residuals show evidence of autocorrelation. It provides two key values, dL and dU, for different sample sizes and model specifications.
Before using the table, identify your sample size (n), number of predictors (k), and chosen significance level (α). Then find the matching critical values for your regression model.
Durbin-Watson Critical Values
| Sample Size (n) | Predictors (k) | Significance Level (α) | Lower Bound (dL) | Upper Bound (dU) |
|---|---|---|---|---|
| [Verified value] | [Verified value] | [Verified value] | [Verified value] | [Verified value] |
| [Verified value] | [Verified value] | [Verified value] | [Verified value] | [Verified value] |
| [Verified value] | [Verified value] | [Verified value] | [Verified value] | [Verified value] |
| [Verified value] | [Verified value] | [Verified value] | [Verified value] | [Verified value] |
| [Verified value] | [Verified value] | [Verified value] | [Verified value] | [Verified value] |
How to read the table: Find your sample size (n) and number of predictors (k). Then select the appropriate significance level (α) to obtain dL and dU.
Your calculated Durbin-Watson statistic is then compared with these critical values. The comparison determines the appropriate test decision or whether the result falls into an inconclusive region.
Important: Critical values can vary with the table convention and test setup. Always verify the source and assumptions behind the values before using them in statistical analysis.
Mobile usability: Keep the table horizontally scrollable on smaller screens. Use clear column labels and avoid unnecessary formatting so readers can quickly find n, k, α, dL, and dU.
How to Use the Durbin-Watson Table
Knowing how to use the Durbin Watson table makes the test much easier. You only need a few values from your regression output and the correct critical-value table.
Follow these steps:
- Find your Durbin-Watson statistic.
Get the statistic from your regression output. It usually appears in the model summary. - Identify your sample size (n).
Count the observations included in the regression model. This value determines which sample-size row you need. - Identify the number of predictors (k).
Count the independent variables included in your regression model. Do not count the intercept as a predictor. - Choose the significance level (α).
Use the significance level specified for your test. Common choices include 0.05 and 0.01. - Find dL and dU.
Locate the row or column matching your n and k values. Then select the appropriate significance level. - Compare your statistic with dL and dU.
Your Durbin-Watson statistic determines which decision region applies. - Determine the appropriate conclusion.
Depending on the comparison, the test may indicate positive autocorrelation, no evidence of positive autocorrelation, or an inconclusive result. The same critical-value framework applies when testing for negative autocorrelation using the corresponding decision boundaries.
Always check the definitions used by your specific critical-values table before making a final conclusion.
What Is k in the Durbin-Watson Table?
In a Durbin-Watson table, k represents the number of explanatory variables, or predictors, in the regression model. It does not represent the total number of variables in your dataset.
For example, suppose your regression model is:
Y = β₀ + β₁X₁ + β₂X₂ + β₃X₃ + ε
This model has three predictors: X₁, X₂, and X₃. Therefore, k = 3.
The intercept, shown as β₀, is not counted as k. This distinction is important because including the intercept can lead you to select the wrong critical values.
A simple rule is:
k = number of independent variables used to explain the dependent variable.
So, if your model contains five predictors and one intercept, use k = 5, not 6.
Before using the table, check how its source defines k. Most standard Durbin-Watson critical-value tables define k as the number of explanatory variables, excluding the constant or intercept.
Getting k correct helps you select the appropriate dL and dU values. A wrong k can lead to an incorrect test decision.
How to Read a Durbin-Watson Table
Learning how to read a Durbin Watson table becomes easier when you follow the same sequence every time. The key is matching your regression details with the correct critical values.
Start by identifying these three values from your regression analysis:
- n: Number of observations in the model
- k: Number of predictors, excluding the intercept
- α: Significance level selected for the test
Then use those values to locate dL and dU in the appropriate critical-values table.
| Step | What to Find | Example |
|---|---|---|
| 1 | Sample size (n) | 50 |
| 2 | Predictors (k) | 3 |
| 3 | Significance level (α) | 0.05 |
| 4 | Lower bound (dL) | [Verified value] |
| 5 | Upper bound (dU) | [Verified value] |
For example, suppose your regression uses 50 observations and 3 predictors. You would first find the section for n = 50 and k = 3.
Next, select the column or section for α = 0.05. The corresponding cells provide dL and dU.
Do not estimate missing values between table entries unless the table’s methodology specifically allows it. Use the exact values provided by your chosen source.
Once you find dL and dU, compare them with your calculated Durbin-Watson statistic. That comparison determines the appropriate decision region.
Durbin-Watson Test Decision Rules
The Durbin-Watson test uses dL and dU to create decision regions. These regions help you assess evidence of positive or negative autocorrelation.
For positive autocorrelation, compare the calculated statistic with dL and dU:
| Comparison | Decision |
|---|---|
| DW < dL | Evidence of positive autocorrelation |
| dL ≤ DW ≤ dU | Inconclusive |
| dU < DW < 4 − dU | No evidence of autocorrelation |
| 4 − dU ≤ DW ≤ 4 − dL | Inconclusive for negative autocorrelation |
| DW > 4 − dL | Evidence of negative autocorrelation |
The middle region requires special attention. A result between dU and 4 − dU does not provide evidence of autocorrelation under this critical-value approach.
The two outer regions indicate evidence in the corresponding direction. The regions near the boundaries remain inconclusive, so you should not force a yes-or-no conclusion.
These rules assume the standard Durbin-Watson critical-value framework. Always use the definitions and assumptions associated with your selected table.
Durbin-Watson Table Example
Suppose you run a regression with 50 observations and 3 predictors. Your selected significance level is α = 0.05, and your regression output gives a Durbin-Watson statistic of 1.85.
Your inputs are:
| Value | Result |
|---|---|
| Sample size (n) | 50 |
| Predictors (k) | 3 |
| Significance level (α) | 0.05 |
| Calculated DW statistic | 1.85 |
| dL | [Verified value] |
| dU | [Verified value] |
First, locate n = 50 and k = 3 in your verified Durbin-Watson table. Then select α = 0.05 to obtain dL and dU.
Next, compare DW = 1.85 with those two critical values. The exact conclusion depends on the verified dL and dU values from your selected table.
For example, if the table gives dU < 1.85 < 4 − dU, the result falls in the region with no evidence of autocorrelation.
If instead 1.85 falls between dL and dU, the result is inconclusive. You should not treat an inconclusive result as proof of autocorrelation or independence.
This example shows why choosing the correct n, k, and α values matters. Even a correct Durbin-Watson statistic can lead to a wrong conclusion if you use the wrong critical values.
Durbin-Watson Statistic vs. Critical Values
The Durbin-Watson statistic and critical values serve different purposes in the test. Understanding this difference helps you avoid incorrect conclusions.
| Term | Meaning |
|---|---|
| DW statistic | Value calculated from the regression residuals |
| dL | Lower critical bound |
| dU | Upper critical bound |
The DW statistic comes directly from your regression analysis. It summarizes the relationship between neighboring residuals.
The dL and dU values come from the appropriate Durbin-Watson critical-values table. You select them based on factors such as n, k, and the significance level.
Think of the DW statistic as your observed result. The dL and dU values provide the reference points needed to evaluate that result.
For example, a DW statistic of 1.7 means little by itself. You still need the appropriate dL and dU values to determine its formal interpretation.
This comparison creates decision regions. Your result may indicate positive autocorrelation, no evidence of autocorrelation, negative autocorrelation, or an inconclusive outcome.
Common Mistakes When Using the Durbin-Watson Table
Small errors can lead to the wrong Durbin-Watson conclusion. Check these common mistakes before interpreting your result.
1. Using the Wrong Sample Size
Use the number of observations included in the regression model. Do not automatically use the number of rows in your original dataset.
2. Using the Wrong k
k represents the number of predictors in the regression model. Do not count the intercept as a predictor.
3. Confusing dL and dU
dL is the lower critical bound, while dU is the upper critical bound. Mixing them can change the decision region.
4. Using the Wrong Significance Level
Make sure the table matches your chosen α level. Using 0.01 values when your test uses 0.05 can produce an incorrect comparison.
5. Treating an Inconclusive Result as Definite
Some DW results fall between critical boundaries. In that situation, the test does not provide a definite conclusion.
6. Assuming the DW Statistic Alone Proves Autocorrelation
A DW statistic below or above 2 only suggests a direction of possible autocorrelation. It does not replace the formal critical-value comparison.
Pro tip: Before interpreting your result, verify n, k, α, dL, and dU. This quick check can prevent many common table-reading errors.
Conclusion
Understanding the Durbin Watson table makes regression analysis easier and helps you evaluate possible autocorrelation. You learned how to identify n, k, and the significance level, then find dL and dU.
You also learned how to read the table, compare critical values with your DW statistic, and understand each decision region. Remember that the DW statistic alone does not prove autocorrelation. The critical-value comparison provides the formal decision framework.
If you’re new to regression analysis, take your time when matching your model details with the table. Checking each value carefully can prevent common interpretation errors.
Found this guide helpful? Share it with someone learning regression analysis, or explore more statistics resources on our website.
Frequently Asked Questions
What is a Durbin-Watson table?
A Durbin-Watson table provides critical values for evaluating autocorrelation in regression residuals. It typically includes dL and dU values based on sample size, predictors, and significance level. You compare your calculated Durbin-Watson statistic with these values to determine the appropriate test decision.
How do you use a Durbin-Watson table?
First, identify your sample size (n) and number of predictors (k). Then select your significance level (α) and find the matching dL and dU values. Compare your calculated Durbin-Watson statistic with these critical values. The comparison determines the appropriate decision region.
How do you read a Durbin-Watson table?
Start by finding your sample size (n) and number of predictors (k). Next, locate the appropriate significance level. The table then provides dL and dU for your model. Compare your DW statistic with these critical values to determine the test outcome.
What is k in a Durbin-Watson table?
In a Durbin-Watson table, k represents the number of predictors in the regression model. It does not include the intercept or constant. For example, a model with four independent variables has k = 4.
What are dL and dU in the Durbin-Watson table?
dL is the lower critical bound, while dU is the upper critical bound. These values help create the decision regions for the Durbin-Watson test. Their values depend on factors such as n, k, and the selected significance level.
What does a Durbin-Watson statistic close to 2 mean?
A Durbin-Watson statistic close to 2 generally suggests little or no first-order autocorrelation in the residuals. However, the statistic alone does not provide a complete hypothesis-test conclusion. For formal testing, compare it with the appropriate critical values.
Team ClassRank builds smart, student-friendly calculators to make grade and rank tracking simple and accurate. For questions or feedback, contact us at classrankcalculator@gmail.com.
