Plot Groups Using PROC SGPLOT in SAS

Method 1: Creating Separate Charts Using the BY Statement

One effective way to manage the visualization of categorized data is to generate individual plots for every unique group. This approach is best suited for situations where the analyst needs to examine the internal characteristics of each group without the distraction of comparison, or when the number of groups is small enough that multiple distinct plots are easy to review. In SAS programming, this functionality is primarily achieved through the use of the BY statement immediately following the PROC SGPLOT call.

The critical functionality of the BY statement is its ability to segment the input data stream based on the values of a specified variable. When PROC SGPLOT encounters a BY statement, it iterates through the data, generating a complete, separate graphical output for every distinct value found in the grouping variable. This ensures that each output chart focuses solely on the observations belonging to one specific group, providing maximum clarity regarding that group’s distribution or trend.

For instance, if we aim to visualize the distribution of ‘points’ scores segmented by ‘team’, the inclusion of `by team;` within the procedure step dictates that SAS will produce one graphical output for Team A, a second graphical output for Team B, and so on. The code snippet below illustrates this mechanism for generating multiple histograms, each dedicated to a single team’s score distribution:

/*create multiple plots that show histogram of points for each team*/
proc sgplot data=my_data; 
   by team;
   histogram points;
   density points / type=kernel;
run;

In-Depth Look at the BY Statement Syntax and Logic

The BY statement is a fundamental component in many SAS procedures, not just PROC SGPLOT. When used, it requires the input dataset to be sorted by the BY variable. Although modern graphical procedures often handle sorting implicitly, it is best practice to ensure the dataset is pre-sorted using PROC SORT to guarantee optimal performance and accurate output, especially when dealing with large datasets. Failure to sort the data might lead to unexpected or incomplete graphical outputs if the data is not ordered correctly.

Within the visualization context, the BY statement acts as a powerful segmentation tool. Every output graph generated by this method will automatically include the BY variable value (e.g., Team A, Team B) in the title or subtitle, providing immediate context for the viewer. This is crucial for maintaining clarity when reviewing dozens of separate charts generated from a dataset containing many different subgroups.

Furthermore, notice the inclusion of the DENSITY statement in the example code: `density points / type=kernel;`. This statement is used to overlay a continuous probability density curve, specifically using kernel density estimation. This curve is non-parametric and serves as an excellent visual summary of the shape of the distribution, smoothing out the noise present in the raw histogram bars. When the BY statement is active, SAS calculates and plots a separate density curve for each individual chart, ensuring the smoothing is relevant only to the data points in that specific group.

Method 2: Overlaying Groups on a Single Chart Using the GROUP Option

The alternative, and often more powerful, method for comparing distributions is plotting all groups onto a single visualization. This technique is indispensable when the primary analytical goal is immediate comparison—understanding how the distributions overlap, where their centers differ, and comparing their respective spreads directly on the same scale. In PROC SGPLOT, this is achieved by utilizing the GROUP= option within the plot statement itself (e.g., HISTOGRAM or SCATTER).

When the GROUP= option is applied, SAS does not segment the output into separate charts. Instead, it maintains a single coordinate system and generates distinct graphical elements for each unique value of the specified grouping variable (e.g., group=team). SAS automatically assigns different visual attributes—such as color, line pattern, or marker symbol—to differentiate the groups, and it creates an appropriate legend to map these attributes back to the group labels.

The resulting single chart provides an efficient platform for comparative analysis. For instance, overlaying two histograms allows the analyst to instantly see which team has a higher median score or whether the variability of scores is greater in one team versus the other. The following code demonstrates how to use the GROUP= option to display separate histograms and their corresponding density curves for each team within a unified plot:

/*plot histogram of points for each team on one chart*/
proc sgplot data=my_data; 
   histogram points / group=team;
   density points / type=kernel group=team;
run;

Comparing the BY Statement and the GROUP Option

Although both the BY statement and the GROUP= option achieve the goal of visualizing data by subgroups, their mechanical implementation and resulting analytical benefits are fundamentally different. Understanding these differences is key to choosing the correct approach for a given visualization task. The BY statement operates at the procedure level, dictating that SAS should execute the entire procedure—including all plot statements—separately for each group, resulting in multiple output graphs.

Conversely, the GROUP= option operates at the plot statement level (e.g., within the HISTOGRAM statement). It instructs the procedure to draw distinct graphical elements corresponding to each group within a single graphical canvas. This is advantageous for direct comparison but can quickly become visually complex if the number of groups is large (e.g., more than 5 or 6 groups) or if the individual group distributions overlap significantly.

When choosing between the two, analysts should consider the volume of groups and the comparison required. If you have many groups (e.g., 20 states), using the BY statement might yield 20 separate, readable charts. If you use GROUP= with 20 groups, the resulting single chart would likely be cluttered and illegible due to color overuse and excessive overlap. For small-scale, direct, side-by-side comparison, the GROUP= option is superior, as it places all distributions on the exact same axis scale, facilitating immediate visual assessment of differences in means and ranges.

Practical Demonstration: Setting Up the Sample Dataset

To illustrate these grouping methods effectively, we will utilize a small, fictional dataset detailing the performance scores (‘points’) for two distinct competitive entities (‘team’). This dataset is constructed using a standard SAS data step, which is the foundational method for creating and initializing datasets within the SAS environment.

The following SAS code block defines the input data, assigning points scores to either Team A or Team B. After the data is created, the PROC PRINT step is used to display the raw data in the output window, allowing verification of the dataset structure before proceeding with graphical analysis.

/*create dataset*/
data my_data;
    input team $ points;
    datalines;
A 29
A 23
A 20
A 21
A 33
A 35
A 31
B 21
B 14
B 15
B 11
B 12
B 10
B 15
;
run;

/*view dataset*/
proc print data=my_data;

The resulting dataset, which is automatically generated and viewed, confirms that the structure contains two variables: team (a character variable denoted by the ‘$’ in the input statement) and points (a numeric variable), ready for visualization.

Example 1: Implementing the BY Statement for Separate Charts

In this first example, we utilize the powerful segmentation capability of the BY statement within PROC SGPLOT. Our objective is to generate two distinct histograms—one detailing the distribution of points for Team A, and the other for Team B. This clear separation ensures that each team’s performance profile is analyzed in isolation, free from visual interference from the other group.

By simply including the line `by team;` immediately after the procedure statement, we instruct SAS to iterate over the dataset, producing a complete histogram plot for every value of the team variable. We also include the DENSITY statement to overlay a smoothed kernel curve on the histogram bars, providing an additional layer of insight into the distributional shape.

/*create multiple plots that show histogram of points for each team*/
proc sgplot data=my_data; 
   by team;
   histogram points;
   density points / type=kernel;
run;

The resulting output confirms that the procedure successfully generated two discrete graphs. The first graph displays the distribution of scores for Team A, which appears to be centered around a higher point value with a moderate spread. The second graph focuses exclusively on Team B, revealing a distribution centered at a lower score, indicating a notable performance gap between the two teams.

The first histogram clearly illustrates the score distribution for team A, while the second separate histogram is dedicated to displaying the distribution of scores for team B.

Note: The optional density statement is highly recommended in distributional plotting, as it overlays a continuous probability curve. This curve provides a smooth, non-binned summary of the data’s distribution shape, which is often more helpful than the stepped bars of the histogram alone, especially when sample sizes are small.

Example 2: Implementing the GROUP Option for Overlaid Charts

For direct visual comparison, we transition to using the GROUP= option within PROC SGPLOT. This method is utilized to create a single, unified chart containing overlaid histograms, allowing the analyst to contrast the distributions of points for each team instantly.

By applying the group=team option to both the HISTOGRAM statement and the DENSITY statement, we instruct SAS to draw the elements for each team onto the same plot, automatically assigning distinct colors and generating a comprehensive legend. To manage the inevitable overlap when distributions are plotted together, we introduce a critical aesthetic option: TRANSPARENCY.

/*plot histogram of points for each team on one chart*/
proc sgplot data=my_data; 
   histogram points / group=team transparency=0.5;
   density points / type=kernel group=team;
run;

The resulting visualization provides an immediate, powerful comparison. The blue histogram displays the distribution of points values for team A and the red histogram displays the distribution of points values for team B. The overlap region and the distinct placement of the density curves make the performance disparity visually striking.

The major benefit of using this overlaid approach is the ability to quickly and efficiently compare the distribution of points values for each team, as both are anchored to the same horizontal axis and scale. This technique is invaluable for rapid analytical insights, especially when assessing whether distributions are sufficiently separated to warrant formal statistical testing.

Enhancing Visualization: Customizing Density and Transparency

When using the GROUP= option, particularly with histograms, the issue of overlap must be addressed to ensure readability. If the bars of different groups completely hide each other, the chart loses its effectiveness. This is where the TRANSPARENCY option becomes essential.

The transparency argument, applied within the plot statement (e.g., `histogram points / group=team transparency=0.5;`), controls the opacity of the graphical elements. The value ranges from 0 (fully opaque) to 1 (fully transparent). By setting this value to 0.5, as demonstrated in Example 2, the histograms become semi-transparent, allowing the viewer to discern the bars of the underlying group where they overlap with the bars of the foreground group.

Effective use of TRANSPARENCY, coupled with the inclusion of the smooth DENSITY plot, transforms a potentially confusing chart into a highly informative comparative visualization. The closer you set this value to 1, the more transparent the histograms become, helping to avoid visual clutter when analyzing highly overlapping distributions.

Further Learning Resources

For those interested in expanding their knowledge of statistical graphics in SAS, the following list includes tutorials on creating other chart types using PROC SGPLOT and related procedures:

• Creating Box Plots: Essential for comparing descriptive statistics (median, quartiles) across multiple groups.
• Generating Scatter Plots: Ideal for visualizing relationships between two continuous variables, often utilizing the GROUP= option to differentiate categorical segments.
• Utilizing PROC GCHART: While PROC SGPLOT is generally preferred for statistical quality, PROC GCHART remains relevant for certain business graphics, such as simple bar charts and pie charts.