Boxplot in SPSS: How to Create and Read the Results

When to use a boxplot

Use a boxplot when you want a quick view of the distribution of a quantitative variable, a comparison of that variable across several groups, or a way to identify observations with unusual values. For example, before running a regression, you can use a boxplot to check whether scores, income, service usage time, or spending are strongly skewed or contain outliers.

A boxplot is suitable for numerical data, such as a variable measured on an interval or ratio scale. If your variable is gender, field of study, or region, SPSS can still use it as a grouping variable on the horizontal axis, but you should not create a boxplot for the categorical variable itself. When you need to check dispersion as well, you can also review standard deviation and variance.

A boxplot does not test a hypothesis and does not automatically show that your data are wrong. It is a visual diagnostic tool. A point beyond the whisker may be a data-entry error, but it may also represent a respondent who genuinely has a very high or very low value. The decision to keep, correct, or remove an observation must be based on the original data and the study design.

In quantitative research, boxplots are commonly used at three points. First, during data cleaning. Second, when checking assumptions before regression, a t-test, or ANOVA. Third, when reporting descriptive statistics so that readers can see differences between groups. If your research question asks whether the value of variable Y differs between group A and group B, the boxplot helps you see the direction of the difference before you read the p-value.

Prepare the data before running the analysis

Open your .sav file and check the variable you want to plot in Variable View. A quantitative variable should have Measure set to Scale. Grouping variables such as gender or income group can be set to Nominal or Ordinal, depending on how they are coded. Also check the variable labels, value labels, and codes such as 99, 999, or -1. If these codes actually mean “no response,” declare them as missing or handle them before creating the plot.

In Data View, each row should represent one respondent and each column should represent one variable. Do not enter text such as “male,” “female,” “high,” or “low” in a column that you plan to average or use for a quantitative boxplot. For a Likert questionnaire, observed variables are usually coded numerically. If you have a reverse-coded variable, process it according to the scale instructions before creating a composite variable. A boxplot for an unprocessed reverse-coded variable can lead you to misread the distribution.

Check the missing values and the number of valid observations. Go to Analyze > Descriptive Statistics > Frequencies, move the variable you want to check into the variable box, and review the frequency table. If a variable has many missing values, the boxplot is calculated only from valid cases, and that number may differ across variables. Do not compare two boxes without checking how many observations each box is based on.

You should also check whether the data were entered using the correct units. An income value entered as 500000 instead of 5.000.000, or a Likert score entered incorrectly as 55, will create an unusual plot. You can also review the variance formula to understand why a few values far from the center can sharply increase dispersion.

If your survey data come from several groups, prepare a grouping variable with clear codes. For example, 1 = Male, 2 = Female. These labels make the chart easier to read and the results easier to write up. Before running the analysis, save a copy of the original data. Every correction or removed observation should be documented because the committee may ask how you cleaned the data.

Steps to create a boxplot in SPSS

Method 1: Create one boxplot for one variable

Go to Graphs > Legacy Dialogs > Boxplot. In the dialog box, select Simple if you want to examine only one variable, then select Summaries of separate variables. Click Define.

Move the quantitative variable you want to check into Boxes Represent. For example, select Diem_hailong. Click OK and SPSS will create the chart in the Output Viewer. This method is suitable when you want to examine the full sample without splitting it into groups.

Method 2: Compare boxplots across groups

Go to Graphs > Legacy Dialogs > Boxplot, select Simple, and then select Summaries for groups of cases. Click Define. Move the quantitative variable into Variable and the grouping variable into Category Axis.

For example, place Diem_hailong in Variable and Gioi_tinh in Category Axis. If you want a clearer title or more readable labels, double-click the chart in the Output Viewer to open Chart Editor, then edit the title and axis labels.

Method 3: Use Chart Builder

Go to Graphs > Chart Builder. If SPSS displays a dialog asking you to confirm the measurement level, choose continue. In the chart gallery, select Boxplot, then drag the simple boxplot template into the preview area.

Drag the quantitative variable to the Y-axis and the grouping variable to the X-axis. If you want one box for the full sample, leave the X-axis empty or use a template designed for one variable. Click OK. Chart Builder is more convenient when you want to adjust several chart elements, but the Legacy Dialogs path is often easier to follow when you are working on a thesis.

Method 4: Display labels for outliers

In an SPSS boxplot, unusual points are usually shown with a case number. Some more distant points may be marked with an asterisk. Double-click the chart, select the data point, or open the chart properties to check its label.

The displayed number is usually the row or case number in the data file, not a student ID or questionnaire ID that SPSS identifies automatically. Match that number against Data View or the respondent ID variable. Do not delete a row simply because it has a symbol on the chart.

Method 5: Save the output and record the processing version

In Output Viewer, select File > Save As to save the output file. You can also right-click the chart and copy it into Word, but keep the .spv file so you can check it later. If you filtered the sample with Data > Select Cases, record the filtering condition in your analysis log.

When you rerun the analysis after correcting the data, name the files by version, such as boxplot_v1.spv and boxplot_v2.spv. This makes it clear which chart was created before and after the outlier processing. It is a small step, but it is useful when the results change between runs.

Read the output

A boxplot does not have a long output table like a regression. You need to read five main components: the median line, the lower edge of the box, the upper edge of the box, the lower whisker, and the upper whisker. The distance from the lower edge to the upper edge is the IQR, or interquartile range. It shows where the central 50% of observations are located.

The line inside the box is the median. If it is near the middle of the box, the central part of the distribution is relatively balanced. If it is closer to one side, the data may be skewed. The taller the box, the more dispersed the central 50% of observations are. When comparing groups, examine the median position and the box height together.

Whiskers are the lines extending from the box to the range of values that SPSS treats as remaining within the usual boxplot range. Points shown separately beyond the whiskers are outliers. An asterisk often indicates a point farther away than points marked with circles, depending on how SPSS displays and numbers cases.

The table below contains illustrative output, not the results of a real study. The numbers simulate the shape you may observe in an SPSS boxplot.

In this example, the median of Group B is higher than the median of Group A. However, you cannot write that the groups differ significantly based only on the visual. To reach that conclusion, run the appropriate test, such as a t-test for two groups or ANOVA for several groups, then read the p-value and check the applicable conditions.

If two boxes overlap, that does not automatically prove that the groups are not different. If two boxes are far apart, that is a useful sign to investigate further. The boxplot gives you a picture of location and dispersion, while the statistical test addresses the evidence for a difference.

You can also read the article about median to distinguish the median from the mean. With skewed data or outliers, the median often describes the center more steadily than the mean. The reported measure must still match the purpose of the analysis and the way you present your study.

Evaluation criteria

A boxplot does not have one “acceptable” threshold like Cronbach's Alpha or KMO. If you see a point beyond the whisker, you should not immediately write that the data fail. The criteria below help you decide what to check next. They do not replace reviewing the observation record and the research model.

In regression, an outlier can affect the coefficient, residuals, and conclusion. After creating the boxplot, you can also check standardized residuals, leverage, or Cook's Distance, depending on the model. These measures should not be replaced by a boxplot because they answer different questions.

When you identify an outlier, a possible sequence is to check data-entry errors, compare the original data, assess whether the response is valid, run the analysis with and without the observation, and report the reason if you remove it. Deleting data just to make the chart look better is difficult to defend. If the observation is valid and belongs to the target population, keeping it often has a stronger basis, but you should describe its effect.

You can read more about outliers to distinguish an unusual point in one variable from an influential point in a model. The two concepts are related, but they are not exactly the same.

How to write this in your thesis

In the results chapter, state which variable was plotted, whether the chart was split by group, and what the chart was used to check. You do not need to describe every point if the chart contains no notable outliers. If there is an unusual point, explain how you checked it and how you decided to handle it.

You can adapt the following sentence to your actual data: “The boxplot shows that the median of [VARIABLE NAME] in [GROUP NAME 1] is [VALUE], while the median in [GROUP NAME 2] is [VALUE]. [GROUP NAME] has [higher/lower dispersion] and contains [NUMBER] observations beyond the whisker. These observations were [retained/removed from the analysis] after [state the basis for checking the data].”

If you have not run the test, avoid writing “Group A is significantly higher.” A safer description is “descriptively, the median of Group A is higher than that of Group B.” Then present the t-test or ANOVA results in the separate testing section.

Make the figure title specific, such as “Figure 4.2. Boxplot of satisfaction scores by gender.” State the measurement unit in the caption if one applies. If you use a composite variable created from several observed variables, explain how the score was calculated, such as the mean of the variables after reliability checking. Do not screenshot the entire Output Viewer if the text is too small. Export the chart at a readable size and keep the original output file in your analysis records.

Common mistakes when creating a boxplot

Selecting the wrong chart type

Some students select a bar chart or histogram and then try to read it as a boxplot. Check the chart icon and name under Graphs > Legacy Dialogs > Boxplot. A histogram shows frequencies by interval, while a boxplot summarizes location, quartiles, and unusual points.

Putting a categorical variable on the Y-axis

If you put gender or field of study into Variable, SPSS may produce a result that is difficult to interpret or may not run as intended. Place the quantitative variable in Variable or on the Y-axis, and place the grouping variable in Category Axis or on the X-axis.

Immediately deleting every point beyond the whisker

A point beyond the whisker is a signal to check, not a delete command. Match the case number, review the original value, and check the collection conditions. If you keep the observation, you can report that the data contain an outlier and explain how you assessed the sensitivity of the result.

Comparing groups with very different sample sizes

You can plot one group with 20 observations and another with 300 observations on the same chart, but the stability of their medians and quartiles will differ. Also report the N for each group in the descriptive statistics table.

Confusing a point number with the respondent ID

The number beside a point on the chart is usually the case number in the file. If the data have been sorted or filtered, it may not match the questionnaire ID. Check the ID variable before contacting a respondent or removing a case.

Drawing a distributional conclusion from the boxplot alone

A boxplot helps identify skewness and outliers, but it does not show the full shape of the distribution. When you need a closer check, also review the histogram, Q-Q plot, and descriptive statistics. For a specific analysis, check that test's assumptions rather than using one chart for every decision.

Frequently asked questions

What is a boxplot?

A boxplot summarizes data using the median, quartiles, whiskers, and points beyond the whisker range. It gives you a quick view of the central location, dispersion, and unusual observations for a numerical variable.

In SPSS, you can create a boxplot for the full sample or split it by group. It is commonly used during descriptive statistics and data checking before inferential analysis.

How should you read a boxplot correctly?

Read the line in the middle of the box as the median, the lower and upper edges as Q1 and Q3, and the distance between those edges as the IQR. The whiskers show the range of values that remain within the chart's usual range. A point separated from a whisker is an observation to check.

When comparing groups, look at the median, box height, whisker length, and number of observations together. The visual shows a descriptive pattern and does not replace the p-value from a statistical test.

What should you do when a boxplot has an outlier?

First, check the case number, entered value, measurement unit, and original data. If it is a valid response, you should not delete it simply to remove the unusual point. You can run the analysis with and without that case to see whether the conclusion is sensitive to the outlier.

If you remove the observation, record the criterion, number, and reason in the data-processing log. If you keep it, report the presence of the outlier and consider descriptive measures suitable for a skewed distribution.

Can you use a boxplot for a Likert variable?

You can use a boxplot for a numerically coded Likert variable or for the mean score of several observed variables, but interpret it carefully because the scale has a limited number of levels. For a multi-item Likert scale, you will often check reliability, calculate the scale score, and then describe the composite variable if that is the procedure for your study.

If you plot each observed variable separately, the chart may contain many tied values and provide limited information. Consider the purpose of the chart before including it in the results chapter.

Can a boxplot replace a t-test or ANOVA?

No. A boxplot shows visual differences between groups, while a t-test or ANOVA tests the difference using a statistical model and its assumptions. You can use the boxplot to identify signs that should be checked before running the test.

If the data violate the conditions of a parametric test, consider an option appropriate for the design and data rather than drawing a conclusion from the visual alone. The final results should include descriptive statistics and a justified statistical test.

Why does SPSS not display the boxplot as expected?

Common causes include text in the Y variable, unprocessed missing-value codes, no valid cases after filtering, or variables placed in the wrong box in the dialog. Check Variable View, the variable frequencies, and the status of Select Cases.

You can also try Graphs > Legacy Dialogs > Boxplot with one simple numerical variable. If the chart runs, add the grouping variable and check each step rather than adding many variables at once.

Open your data file, select one quantitative variable, and create a boxplot through Graphs > Legacy Dialogs > Boxplot. Then save the chart, output file, and notes on your decision for each outlier. If you need to run the analysis on your own .sav or .csv file, use M4 data analysis.