A Comprehensive Guide to Generating Summary Statistics in SAS with PROC SUMMARY and the NWAY Statement

In the realm of statistical computing, particularly when leveraging the robust capabilities of SAS, data analysts frequently require the generation of concise and highly targeted summary statistics. The primary tool for this within the SAS environment is the PROC SUMMARY procedure, which efficiently aggregates large volumes of data and calculates essential descriptive measures. A frequent challenge arises because `PROC SUMMARY`’s default operation, known as “rollup” processing, yields a comprehensive, multi-level output. This output invariably includes statistics for the overall population in addition to the desired specific group level summaries. When the analytical goal is strictly limited to obtaining metrics at the most granular level of grouping, this excess information can significantly clutter the report, making interpretation challenging. The powerful, yet remarkably simple, statement, NWAY, offers an indispensable solution to this complexity.

When integrated into the PROC SUMMARY syntax, the NWAY statement acts as a precise filter for the analytical output. By invoking NWAY, the SAS system is specifically directed to compute and display results only for the combinations defined by the specified class variables. This action systematically filters out all higher-level aggregations and overall totals, resulting in a dramatically cleaner, more focused report that addresses the exact group levels of interest, thereby significantly enhancing both the readability and interpretability of the summary results. This comprehensive guide will detail the practical application of the NWAY statement within standard statistical workflows. We will use a concrete, step-by-step example, contrasting the verbose default output generated by PROC SUMMARY with the streamlined results achieved when NWAY is properly employed. Mastering this fundamental utility is essential for efficient data analysis and professional reporting within the SAS environment.

Exploring the SASHELP.FISH Dataset

To effectively demonstrate the precise filtering capabilities afforded by the NWAY option, we will utilize a foundational, built-in SAS dataset called SASHELP.FISH. This dataset is widely recognized and frequently used as a standard teaching resource in SAS documentation and training modules, making it an ideal candidate for illustrating complex statistical and data manipulation procedures. The SASHELP.FISH dataset contains detailed biological measurements for 159 individual fish specimens collected from a lake in Finland, representing a diverse range of species. Each observation within the dataset includes several critical variables, such as the fish’s species, various length measurements, and its weight. This rich categorical and quantitative structure makes it perfectly suited for generating group-level summary statistics based on variables like species.

Before proceeding with any form of aggregation or summary calculation using PROC SUMMARY, best practice in data analysis dictates a preliminary inspection to confirm the data’s structure and content. We employ the PROC PRINT procedure to display the initial observations, providing a rapid overview of the raw data we intend to analyze.

The following SAS code snippet executes this crucial preliminary step, displaying the first 10 observations extracted from the SASHELP.FISH dataset:

/*view first 10 observations from Fish dataset*/
proc print data=sashelp.Fish (obs=10);

run;

This initial examination confirms the presence and naming conventions of key variables, such as Species and Weight. Gaining this deep understanding of the source data is a mandatory first step before performing any meaningful aggregation or generating summary statistics.

Default Behavior of PROC SUMMARY Without NWAY

When a user executes PROC SUMMARY without explicitly including the NWAY statement, the procedure is designed to produce descriptive statistics across all possible hierarchical combinations of the CLASS statement variables. Crucially, it also computes statistics for the entire dataset, which represents the highest level of aggregation. This default multi-level output, often referred to as a “rollup,” provides a comprehensive, albeit frequently overly detailed, perspective on the underlying data structure. For our illustration, our primary objective is singular: to calculate descriptive statistics specifically for the Weight variable, categorized and grouped distinctly by the Species variable. In the absence of the NWAY option, the resulting output will contain not only the statistics for each unique species (our desired output) but also statistics representing the overall aggregate data, combining all species together in a single summary.

Examine the following SAS code. It invokes PROC SUMMARY to calculate standard summary statistics for Weight, categorized by Species. The OUTPUT OUT=summaryWeight; statement is essential, as it directs the calculated statistics into a new output dataset named summaryWeight. Subsequently, the PROC PRINT procedure is used to display the contents of this newly created output dataset.

/*calculate descriptive statistics for Weight, grouped by Species*/
proc summary data=sashelp.Fish;
    var Weight;
    class Species;  
    output out=summaryWeight;
run;

/*print output dataset*/
proc print data=summaryWeight;

Although the screenshot above only displays the initial 20 rows, the full output dataset, summaryWeight, actually contains 40 distinct rows of data due to the hierarchical rollup. This visual snippet is sufficient, however, to clearly demonstrate the inherent structure and comprehensive content of the default, multi-level output produced by PROC SUMMARY when the NWAY statement is intentionally omitted.

Deciphering the Comprehensive Output Structure

The standard output produced by PROC SUMMARY, when unrestricted by NWAY, delivers a rich array of information, meticulously organized into several key, automatically generated columns. These columns are essential metadata required for decoding the various aggregation levels present in the rollup. Understanding the precise role of each column is vital for accurate interpretation of the results:

• _TYPE_: This unique, automatically generated variable is included by `PROC SUMMARY` to signify the specific level of aggregation represented by that row. A value of 0 invariably indicates summary statistics calculated for the entire dataset—the overall total. Progressively higher values correspond to summaries based on different combinations of the CLASS statement variables. The maximum value (1 in this case, as we only used one grouping variable, Species) represents the most detailed group level, where all defined CLASS statement variables are utilized for grouping.
• _FREQ_: This column reports the number of observations or records that contributed to the calculation of the descriptive statistics for that particular group level or overall summary. Essentially, it functions as the frequency count of the records used in the calculation.
• _STAT_: This variable clearly labels the specific descriptive statistic being reported in the corresponding row. Standard values reported by default include ‘N’ (the count of non-missing observations), ‘MIN’ (the minimum value), ‘MAX’ (the maximum value), ‘MEAN’ (the arithmetic mean), and ‘STD’ (the standard deviation).
• Weight: Finally, this column presents the actual numerical result for the descriptive statistic indicated in the _STAT_ column, specifically for the Weight variable under analysis.

Upon closer inspection of the default output, the initial rows, which all show _TYPE_ = 0, provide the summary statistics for the entire SASHELP.FISH dataset, completely disregarding the grouping by Species. For instance, these overall rows immediately tell us that the total count of non-missing observations (N) for Weight was 158, the minimum recorded weight was 0, and the mean weight across all fish was approximately 398.70. Following these comprehensive, overall statistics, the subsequent rows then display the identical set of descriptive statistics, but meticulously calculated for each unique species, starting with ‘Bream’ and continuing for all other defined categories. This tiered, hierarchical presentation is the fundamental default behavior of PROC SUMMARY.

Introducing the NWAY Statement for Focused Summaries

While the default hierarchical output of PROC SUMMARY is undoubtedly thorough, it frequently includes aggregate summary levels—like the overall total—that are superfluous or entirely irrelevant to a specific analytical question. If the sole purpose of an analysis is to conduct a direct comparison of summary statistics among the distinct groups defined by the CLASS statement variables, the inclusion of overall total statistics (the rollup) simply introduces unnecessary complexity and visual noise. The NWAY statement is deployed precisely to mitigate this issue.

This powerful option, utilized within `PROC SUMMARY`, explicitly instructs the SAS procedure to restrict the aggregation process. It limits the output to only the “N-way” combinations of the specified CLASS statement variables. Operationally, this means that only the rows corresponding to the highest possible value in the _TYPE_ column of the output will be generated and displayed, thus eliminating the less granular summaries.

Considering our current example, which involves grouping by a single CLASS statement variable (Species), the highest _TYPE_ value achievable is 1. Consequently, by activating NWAY, we ensure that SAS produces summary statistics solely for each individual species. This crucial filtering step completely eliminates the initial rows that represented the overall dataset summary (where _TYPE_ = 0), resulting in a direct, concise, and focused report centered exclusively on inter-species comparisons.

Applying NWAY for Targeted Group Analysis

Integrating the NWAY statement into the PROC SUMMARY procedure is remarkably straightforward, requiring only the addition of the keyword NWAY directly onto the `PROC SUMMARY` statement itself. Despite this minimal change in syntax, the modification profoundly transforms the output structure, making it perfectly suited for analyses that require only detailed group-level summary statistics. Let us re-examine our running example using the SASHELP.FISH dataset. We proceed to calculate descriptive statistics for the Weight variable, grouped by Species. This time, however, the critical NWAY option is explicitly included in the procedure call. The subsequent PROC PRINT statement will then display the contents of the refined output dataset, summaryWeight.

Note carefully the placement of NWAY within the `PROC SUMMARY` statement in the code provided below:

/*calculate descriptive statistics for Weight, grouped by Species*/
proc summary data=sashelp.Fish nway;
    var Weight;
    class Species;  
    output out=summaryWeight;
run;

/*print output dataset*/
proc print data=summaryWeight;

Reviewing the output generated with the NWAY statement reveals a stark and immediate difference: the summary statistics corresponding to the entire dataset (all rows where _TYPE_ = 0) are completely absent. Instead, the output dataset is now streamlined, containing only the summary statistics calculated individually for each unique Species. This results in an output that is significantly more manageable, concise, and directly relevant when the focus of the analysis is solely on comparing specific group levels against one another. The effective use of NWAY drastically reduces the verbosity inherent in the default `PROC SUMMARY` output, facilitating easier identification and interpretation of the statistics pertinent to your CLASS statement variables. This targeted approach is particularly advantageous in scenarios involving complex, multi-variable groupings, automated reporting requirements, or when dealing with extremely large datasets where comprehensive rollups would be overwhelmingly cumbersome.

Conclusion: Mastering NWAY for Streamlined Reporting

The NWAY statement stands as an indispensable feature within PROC SUMMARY, providing every SAS user with the ability to precisely refine and focus their statistical output. By ensuring that only the summary statistics corresponding to the most granular group level, as defined by the CLASS statement variables, are displayed, NWAY successfully eliminates extraneous rollup statistics. The outcome is cleaner, more efficient, and far more readily interpretable results.

This powerful, targeted aggregation capability is especially beneficial in analytical contexts where precise, direct comparisons between categories are paramount, or when the outputs from `PROC SUMMARY` must be seamlessly integrated into subsequent automated reports or downstream analytical steps. Proficiency in utilizing NWAY empowers the analyst to produce highly focused and relevant descriptive statistics, significantly enhancing the overall clarity and measurable impact of their data analysis within the SAS ecosystem. By mastering the strategic deployment of NWAY, you gain the assurance that your PROC SUMMARY results are not only accurate but also optimally structured for your specific analytical requirements, thereby saving considerable time and substantially improving the overall efficiency of your statistical workflows.

Additional Resources

To further enhance your technical proficiency in SAS programming and statistical analysis, we recommend exploring the following related tutorials that explain how to perform other common data manipulation and descriptive tasks:

• How to Calculate Mean in SAS
• How to Calculate Standard Deviation in SAS
• Introduction to PROC PRINT in SAS