How to use the NMISS Function in SAS (With Example)

You can use the NMISS functionality in SAS to count the number of missing values for each numeric variable in a dataset efficiently.

Here is one common way to execute this function in practice:

proc means data=my_data nmiss;
run;

This particular example will count the number of missing values for each numeric variable in the dataset called my_data, producing a summary table critical for initiating data cleaning procedures.

Practical Demonstration: Constructing a Dataset with Intentional Missingness

To fully illustrate the mechanism of the NMISS function, we will establish a sample dataset detailing statistics for various basketball players. This dataset, designated my_data, includes a mix of numeric (points, assists, rebounds) and character variables (team) and is intentionally populated with several missing values, represented by a period in the input stream, to mimic typical real-world data imperfections.

The process initiates with the DATA step to define the dataset structure and variable types, followed by the DATALINES statement to input the raw data records. This setup is paramount for ensuring that SAS correctly interprets the variable attributes. Special attention is given to the input of missing numerical data using the period (`.`) symbol, which SAS recognizes as a system missing value. Note that one row also features a missing value for the character variable, further complicating the data structure.

Following data creation, the PROC PRINT statement is executed to display the raw contents of the dataset. This step visually confirms the successful compilation of the data table and permits a manual, preliminary verification of the pattern and location of the missing entries before the NMISS analysis is run.

Example: Using NMISS in SAS to Count Number of Missing Values for Numeric Variables

Suppose we have the following dataset in SAS called my_data that contains information about various basketball players:

/*create dataset*/
data my_data;
    input team $ points assists rebounds;
    datalines;
A 10 2 .
A 17 5 .
A 17 . .
A 18 3 4
A 15 0 5
B . 4 5
B 29 0 8
B . 2 9
C 12 1 9
. 30 1 .
;
run;

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

It is readily apparent that missing values (indicated by the period) are present across the points, assists, and rebounds numeric variable columns, alongside a missing observation in the categorical team column.

Executing the NMISS Count via PROC MEANS

Following the successful creation of the sample dataset, the subsequent phase in the data cleaning methodology involves precisely quantifying the extent of numeric missingness. As previously detailed, this quantification is optimally performed using the PROC MEANS procedure augmented by the NMISS option. This combined command efficiently executes the internal NMISS counting logic across all eligible numeric variables (points, assists, and rebounds), consolidating the results into a concise and easily interpretable summary table.

Crucially, the character variable team is automatically excluded from this PROC MEANS analysis, as the procedure is fundamentally geared towards statistical measures pertinent to continuous data. The structure of the output table is designed to be highly informative, providing the necessary statistical metrics that allow analysts to assess not only the count of missing records but also the reliability and robustness of the calculated averages and distributions for each metric.

We execute the following code to apply the NMISS function through PROC MEANS, thereby calculating the total number of missing observations per numeric variable:

/*count number of missing values in each variable*/
proc means data=my_data nmiss;
run;

Analyzing the Procedural Output and Missing Data Patterns

The output table generated by PROC MEANS serves as quantified evidence of the data deficiencies within the numeric fields. Each row in the table corresponds to one variable, and the essential column labeled NMISS displays the precise count of missing values determined by the NMISS logic. The column labeled N is equally important, representing the count of non-missing, or available, observations that were used in the calculation of summary statistics like the mean.

Reviewing this data allows analysts to draw definitive conclusions about the completeness of their numerical measures, informing subsequent strategic decisions. If the NMISS count for a variable is disproportionately large compared to N, analysts may consider the variable too sparse for meaningful inclusion in predictive models, potentially justifying variable exclusion or the use of more sophisticated, resource-intensive imputation techniques to salvage data utility.

Based on the quantitative summary derived from our example table, we can accurately detail the missingness across the dataset:

• The points variable exhibits 2 missing values. This signifies that 8 out of 10 observations were available for calculating central tendency and dispersion measures.

• The assists variable shows only 1 missing value. This variable is highly complete, which suggests high confidence in its calculated summary statistics.

• The rebounds variable records 4 missing values. This represents the highest rate of missingness in the dataset, necessitating careful consideration during the data cleaning phase.

Counting Missing Character Variables Using PROC SQL

A significant limitation of the default NMISS function, particularly when integrated with PROC MEANS, is its inherent inability to process character variables. In SAS, a missing character value is defined as a field containing only blank spaces or an empty string, which is conceptually distinct from the numeric system missing value (`.`). To accurately extend the missingness assessment to character fields, expert analysts utilize PROC SQL, which facilitates the direct application of functions within a highly flexible query structure.

Within the PROC SQL environment, the NMISS function can be applied to character variables. In this specific context, NMISS evaluates whether the variable’s value conforms to the definition of a missing character string. This method is indispensable for ensuring a holistic data quality assessment, encompassing organizational codes, textual identifiers, and categorical data fields, which are just as susceptible to missingness errors as numerical measurements.

The following example code illustrates the targeted use of the NMISS function on the team variable—a character field—to determine its missing observation count:

/*count number of missing values for team variable*/
proc sql; 
    select nmiss(team) as missing_team_values
    from my_data;
quit;

The output generated by the PROC SQL query confirms that there is exactly 1 missing value recorded in the team column. This concludes the comprehensive assessment of missing data across both numeric and character data types within the sample dataset.

Advanced Considerations and Alternatives

While the NMISS function is highly effective for basic counting, the SAS system provides additional specialized functions and procedures for sophisticated data validation and quality control, particularly when analysts must address complex missing data patterns or implement advanced imputation models. A notable alternative is the CMISS function, often used within the DATA step, which is designed to count missing values across both numeric variables and character variables simultaneously when they are explicitly listed as arguments.

For research that requires detailed pattern analysis—such as identifying whether data is Missing At Random (MAR) or Missing Completely At Random (MCAR)—procedures like PROC MI (Multiple Imputation) or PROC FREQ offer deeper diagnostics. These advanced procedures move beyond simple counting to analyze the type and underlying mechanism of missing data, which is essential for determining the most appropriate statistical modeling approach. Nevertheless, the rapid, foundational count provided by NMISS remains the quickest method for preliminary assessment of numeric data integrity.

In conclusion, mastering the use of NMISS, whether integrated into PROC MEANS for efficiency or utilized within PROC SQL for character-specific assessment, is a non-negotiable step in any professional statistical workflow. This simple yet powerful tool ensures efficient data cleaning and establishes a defensible foundation for subsequent statistical analysis.

The following tutorials explain how to perform other common tasks in SAS: