Dot-and-Whisker Plot
A coefficient plot that displays regression estimates as dots with confidence interval whiskers — making it easy to see which predictors are significant and how they compare in magnitude.
// 01 — The chart
What it looks like
A dot-and-whisker plot showing six regression coefficients. Each dot represents a coefficient estimate and whiskers show the 95% confidence interval. The dashed line at zero marks no effect — whiskers crossing this line indicate non-significance.
// 02 — Definition
What is a dot-and-whisker plot?
A dot-and-whisker plot (also called a coefficient plot or coefplot) is a statistical visualization that displays the estimated coefficients from a regression model. Each predictor variable is represented by a single row containing a dot (the point estimate) and a horizontal line (the confidence interval, or “whisker”).
The key reference element is a vertical line at zero. If a coefficient’s whisker crosses zero, the predictor is not statistically significant at the chosen confidence level. Coefficients entirely to the left of zero have a negative association with the outcome, while those entirely to the right have a positive association.
Unlike a regression table, which requires the reader to mentally parse rows of numbers, a dot-and-whisker plot offers an instant visual comparison of effect sizes and their uncertainty. You can immediately see which predictors matter most and how precisely their effects are estimated.
Origin: The coefficient plot emerged from the political science and social science communities in the 2000s as researchers sought more intuitive ways to present regression results. Kastellec and Leoni’s influential 2007 paper “Using Graphs Instead of Tables in Political Science” advocated for these visual summaries over dense numerical tables.
// 03 — Anatomy
Parts of a dot-and-whisker plot
// 04 — Usage
When to use it — and when not to
- Presenting results from a linear, logistic, or other regression model
- You want to compare the relative importance of multiple predictors
- Showing which coefficients are statistically significant at a glance
- Comparing model results side by side (e.g., Model 1 vs Model 2)
- Replacing a dense regression table for a presentation or paper
- Your audience understands basic regression output and confidence intervals
- You have only one predictor — a simple text summary suffices
- Your model has 30+ predictors, making the plot too dense to read
- Coefficients are on wildly different scales and you haven't standardized
- You need to show non-linear relationships — use partial dependence plots
- Your audience is unfamiliar with regression — use simpler visuals
- You want to show predictions rather than model parameters
// 05 — Reading guide
How to read a dot-and-whisker plot
Follow these steps whenever you encounter a coefficient plot in the wild.
Locate the zero line
The vertical dashed line at zero is your reference. Everything to the left indicates a negative association with the outcome; everything to the right indicates a positive association.
Check each dot's position
The dot is the coefficient's point estimate. Its distance from zero tells you the magnitude of the effect. Farther from zero means a stronger effect on the outcome variable.
Examine the whisker width
Wider whiskers mean more uncertainty about the true coefficient value. Narrow whiskers indicate a precise estimate. Compare whisker widths across predictors to gauge relative precision.
Test for significance visually
If a whisker crosses the zero line, the coefficient is not statistically significant at the chosen confidence level. If the entire whisker sits on one side of zero, the effect is significant.
Compare across predictors
Look at relative dot positions to judge which predictors have the strongest effects. Remember that coefficients are only directly comparable when variables are on the same scale (e.g., standardized).
// 06 — Pitfalls
Common mistakes
Comparing unstandardized coefficients across different scales
Fix: If one predictor is in dollars and another is binary (0/1), their coefficients are not directly comparable. Standardize coefficients or use a common scale to make visual comparisons meaningful.
Including the intercept in the plot
Fix: The intercept is on a fundamentally different scale than slope coefficients and can distort the axis. Exclude it from the plot or show it separately.
Using only 90% confidence intervals to inflate significance
Fix: Always state the confidence level clearly. The standard is 95% CI. Switching to 90% without disclosure is misleading and inflates the appearance of significant effects.
Plotting too many predictors without grouping
Fix: A plot with 40 coefficients becomes unreadable. Group predictors by category (demographics, financial, behavioral) or use facets to keep each panel manageable.
Ignoring multicollinearity
Fix: Correlated predictors inflate confidence intervals and can flip coefficient signs. Check variance inflation factors (VIF) before interpreting the plot, as wide whiskers may reflect collinearity rather than true uncertainty.
// 07 — In the wild
Real-world examples
Political science research
Dot-and-whisker plots are the dominant way to present regression results in top political science journals like the American Political Science Review. Researchers use them to show how factors like income, education, and party affiliation predict voting behavior, often comparing models with and without control variables side by side.
Epidemiological risk factor analysis
In public health studies, coefficient plots display adjusted odds ratios or hazard ratios from multivariable models. For example, a study on cardiovascular disease risk might plot the effects of smoking, BMI, age, blood pressure, and cholesterol simultaneously, making it easy to see which risk factors have the strongest independent association.
Machine learning model interpretation
Data scientists use coefficient plots to interpret regularized regression models like LASSO or ridge regression. By visualizing which features survive regularization (non-zero coefficients), teams can quickly identify the most important predictors in models with hundreds of candidate features.
// 08 — Quick reference
Key facts
// 09 — Variations
Types of dot-and-whisker plots
The basic pattern adapts to different modeling contexts while preserving the core dot-plus-whisker structure.
Standard coefficient plot
The classic single-model format. One dot per predictor with 95% CI whiskers and a zero reference line.
Multi-model comparison
Shows coefficients from two or more models side by side using different colors. Reveals how estimates change when controls are added.
Thick-thin whisker plot
Uses a thick inner whisker for the 50% or 80% CI and a thin outer whisker for the 95% CI, conveying two levels of uncertainty simultaneously.
Weighted coefficient plot
Dot size encodes an additional variable such as sample size or variable importance, adding a second dimension to the standard format.
// 10 — FAQs
Frequently asked questions
What is a dot-and-whisker plot?+
A dot-and-whisker plot (also called a coefficient plot or coefplot) is a statistical visualization that displays the estimated coefficients from a regression model. Each predictor variable is represented by a single row containing a dot (the point estimate) and a horizontal line (the confidence interval, or "whisker").
When should you use a dot-and-whisker plot?+
Use a dot-and-whisker plot when presenting results from a linear, logistic, or other regression model. It also works well when you want to compare the relative importance of multiple predictors, and when showing which coefficients are statistically significant at a glance.
When should you avoid a dot-and-whisker plot?+
Avoid a dot-and-whisker plot when you have only one predictor — a simple text summary suffices. It is also a poor fit when your model has 30+ predictors, making the plot too dense to read, or when coefficients are on wildly different scales and you haven't standardized.
Is a dot-and-whisker plot suitable for dashboards?+
Yes — a dot-and-whisker plot can work well in dashboards as long as the panel is large enough for readers to perceive the encoded values, has a clear title, and includes the legend or axis labels needed to interpret it.
What category of chart is a dot-and-whisker plot?+
Dot-and-Whisker Plot belongs to the Statistical family of charts. Charts in that family are designed to answer the same kind of question, so they often work as alternatives when one doesn't quite fit your data.
How do you read a dot-and-whisker plot?+
Start with the axis labels and legend, then look at the overall shape before zooming into individual marks. Compare prominent features against the rest of the data, and verify any conclusion against the underlying numbers when precision matters.