*Copyright @ www.mycsg.in;

Introduction to SAS - Part 06: Variable Labels, Dataset Labels, and Display Formats

Overview

SAS stores variable names in a compact, code-friendly form — typically short uppercase names such as SUBJ, TRTP, or AVAL.
These short names are efficient for programming but are not always easy to read in reports and output.
SAS allows you to attach a label to any variable or dataset. The label is a longer, descriptive piece of text that SAS displays in output instead of the raw name.
Similarly, a format tells SAS how to display the value of a variable — for example showing 1 as Male and 2 as Female, or displaying a date number as 01JAN2024.
Labels and formats are stored as metadata inside the dataset. They travel with the data and are applied automatically whenever the dataset is used in a PROC step that supports them.
This lesson covers how to assign variable labels, dataset labels, and display formats, and how to confirm they have been stored correctly.

Variable Labels

A variable label is a text string of up to 256 characters that describes what a variable contains.
Labels are assigned using the LABEL statement inside a DATA step or inside PROC steps that support it such as PROC DATASETS.
Once assigned, the label is stored permanently in the dataset descriptor and will appear in PROC PRINT, PROC MEANS, PROC FREQ, PROC REPORT, and other output-producing procedures.
The variable name itself does not change — the label is purely a display attribute.

*------------------------------------------------------------------------------;
* Create a small dataset and assign variable labels in the DATA step ;
*------------------------------------------------------------------------------;
data work.subjects;
    input subj $ age sex $ trt $;
    label
        subj = "Subject Identifier"
        age  = "Age at Screening (years)"
        sex  = "Sex"
        trt  = "Treatment Group";
    datalines;
001 34 M A
002 28 F B
003 45 M A
004 52 F B
;
run;

The LABEL statement lists each variable name followed by an equals sign and the label text in quotes.
Multiple variables can be labelled in a single LABEL statement — each assignment is separated by a space or new line.
Labels are saved into the dataset as part of its descriptor. You do not need to re-assign them every time you use the dataset.

Confirming Labels with PROC CONTENTS OUT=

PROC CONTENTS describes the structure of a dataset — variable names, types, lengths, formats, and labels.
The OUT= option saves that descriptor information to a SAS dataset instead of only printing it to the output tab.
Saving to a dataset lets us inspect the label and format columns directly in a data view, which is more convenient than reading the output tab.
Key columns in the OUT= dataset: NAME (variable name), LABEL (assigned label text), FORMAT (assigned format), TYPE (1=numeric, 2=character), LENGTH.

*------------------------------------------------------------------------------;
* Capture the dataset descriptor to a dataset using OUT= ;
* NOPRINT suppresses the output tab listing since we will view the dataset ;
*------------------------------------------------------------------------------;
proc contents data=work.subjects out=work.contents_subjects noprint;
run;

NOPRINT tells SAS not to write to the output tab — the descriptor information goes only to the OUT= dataset.
Inspect CONTENTS_SUBJECTS in the data view below and locate the LABEL column — confirm that each variable shows the label text assigned in the DATA step.
The NAME column holds the variable name and the LABEL column holds the descriptive text. A blank LABEL means no label has been assigned for that variable.

Dataset Labels

As well as labelling individual variables, you can attach a label to the dataset itself.
A dataset label is assigned using the LABEL= dataset option in the DATA statement.
The dataset label is stored in the descriptor and is visible in PROC DATASETS library listings, making it easy to identify a file's purpose when browsing a library with many datasets.

*------------------------------------------------------------------------------;
* Assign a dataset label using the LABEL= dataset option ;
*------------------------------------------------------------------------------;
data work.subjects (label="Screening Subjects - All Enrolled");
    set work.subjects;
run;
 
*------------------------------------------------------------------------------;
* Capture the updated descriptor to confirm the dataset label was stored ;
*------------------------------------------------------------------------------;
proc contents data=work.subjects out=work.contents_subjects2 noprint;
run;

The LABEL= option appears in parentheses immediately after the dataset name — this is a dataset option, distinct from the variable LABEL statement.
In the CONTENTS_SUBJECTS2 data view below look for the MEMLABEL column — this holds the dataset-level label text.
Variable-level labels are in the LABEL column as before; the dataset label is stored separately in MEMLABEL.
Dataset labels are especially useful in permanent libraries where many datasets are stored and readers need to understand each file's purpose at a glance.

Assigning Labels with PROC DATASETS

You do not always need to re-run a DATA step to add or change labels. PROC DATASETS can modify the descriptor of an existing dataset without reading or rewriting the data.
This is much faster for large datasets because no observations are processed — only the metadata is updated.
Use the MODIFY statement followed by a LABEL statement inside PROC DATASETS to assign or change variable labels.

*------------------------------------------------------------------------------;
* Update a variable label without re-running the DATA step ;
*------------------------------------------------------------------------------;
proc datasets library=work nolist;
    modify subjects;
        label age = "Age at Baseline (years)";
quit;
 
*------------------------------------------------------------------------------;
* Confirm the label change was stored ;
*------------------------------------------------------------------------------;
proc contents data=work.subjects out=work.contents_subjects3 noprint;
run;

MODIFY subjects tells PROC DATASETS which dataset to update — no observations are read or written, only the descriptor changes.
The LABEL statement inside PROC DATASETS works exactly as in a DATA step — variable name equals label text.
Inspect CONTENTS_SUBJECTS3 below and confirm that the LABEL column for AGE now reads Age at Baseline (years) rather than the original text.
This approach is the preferred way to correct a label on a large permanent dataset without the overhead of a full DATA step pass.

Display Formats

A format tells SAS how to display a variable's value in output. The stored value is unchanged — only the displayed representation changes.
SAS has many built-in formats. Common examples include DATE9. for displaying date values as 01JAN2024, DOLLAR12.2 for currency, and $UPCASE. for uppercase character display.
Formats are assigned using the FORMAT statement inside a DATA step, or via PROC DATASETS without rewriting data.
Once assigned, the format name is stored in the dataset descriptor. We can confirm it using PROC CONTENTS OUT= and inspecting the FORMAT column.

*------------------------------------------------------------------------------;
* Create a dataset with a date variable and assign a display format ;
*------------------------------------------------------------------------------;
data work.visits;
    input subj $ visitdt :date9. visitnum;
    format visitdt date9.;
    label
        subj     = "Subject Identifier"
        visitdt  = "Visit Date"
        visitnum = "Visit Number";
    datalines;
001 01JAN2024 1
001 15JAN2024 2
002 03JAN2024 1
002 17JAN2024 2
;
run;
 
*------------------------------------------------------------------------------;
* Confirm the format name is stored in the descriptor ;
*------------------------------------------------------------------------------;
proc contents data=work.visits out=work.contents_visits noprint;
run;

The FORMAT statement assigns the DATE9. format to VISITDT. Without this format the date would display as a raw integer — the number of days since 01JAN1960.
The INFORMAT :DATE9. in the INPUT statement tells SAS how to read the raw text into a date value. The FORMAT statement controls how that value is displayed.
Inspect CONTENTS_VISITS below and locate the FORMAT column — it should show DATE9 next to VISITDT, confirming the format was stored in the descriptor.
Inspect the VISITS data view — the VISITDT column will display as a formatted date because the format is attached to the variable.

Assigning Formats with PROC DATASETS

Just like labels, formats can be added or changed on an existing dataset using PROC DATASETS without rewriting the data.
Use the FORMAT statement inside a MODIFY block to assign a format. To remove a format write FORMAT varname; with no format name — SAS clears the stored format.

*------------------------------------------------------------------------------;
* Assign a format to an existing variable using PROC DATASETS ;
*------------------------------------------------------------------------------;
proc datasets library=work nolist;
    modify subjects;
        format sex $upcase.;
quit;
 
*------------------------------------------------------------------------------;
* Confirm the format name is now stored in the descriptor ;
*------------------------------------------------------------------------------;
proc contents data=work.subjects out=work.contents_subjects4 noprint;
run;

The FORMAT statement inside MODIFY works the same way as in a DATA step — variable name followed by the format name.
Inspect CONTENTS_SUBJECTS4 below and confirm the FORMAT column for SEX now shows $UPCASE, confirming the format was attached without a DATA step rewrite.
To remove a format entirely use FORMAT sex; with no format name in the MODIFY block. The FORMAT column will be blank in PROC CONTENTS after that.

The ATTRIB Statement

The ATTRIB statement is a convenient way to assign a label, format, informat, and length for a variable all in one place.
Using ATTRIB keeps variable attribute definitions together and avoids having separate LABEL, FORMAT, and LENGTH statements scattered through the DATA step.
Each variable gets its own ATTRIB clause and any attributes not specified are left unchanged or take their default values.

*------------------------------------------------------------------------------;
* Use ATTRIB to assign label, format, and length together ;
*------------------------------------------------------------------------------;
data work.summary;
    attrib
        subj    length=$3   label="Subject Identifier"
        age     length=8    label="Age at Screening (years)"
        aval    length=8    label="Analysis Value"          format=8.1
        trtp    length=$20  label="Planned Treatment"
        adtc    length=$10  label="Analysis Date (ISO)";
    input subj $ age aval trtp $ adtc $;
    datalines;
001 34 102.5 DrugA 2024-01-01
002 28  98.0 PlacB 2024-01-03
003 45 115.3 DrugA 2024-01-02
;
run;
 
*------------------------------------------------------------------------------;
* Confirm that all labels and formats were stored from the ATTRIB statement ;
*------------------------------------------------------------------------------;
proc contents data=work.summary out=work.contents_summary noprint;
run;

Each ATTRIB clause names the variable followed by one or more attribute keywords: LENGTH=, LABEL=, FORMAT=, INFORMAT=.
All attributes for a variable are grouped on one line making the data structure easy to read and maintain.
The ATTRIB statement is especially common in clinical programming where datasets have many variables each requiring a label, length, and sometimes a format.
Inspect CONTENTS_SUMMARY below — the LABEL column should show the assigned text and the FORMAT column should show 8.1 for AVAL. All other variables should have blank FORMAT since none was specified.

Key Points

Variable labels are descriptive text attached to a variable using the LABEL statement. They are stored in the dataset descriptor and appear as column headings in output procedures.
Dataset labels are assigned using the LABEL= dataset option and identify the dataset's purpose in library listings.
Formats control how values are displayed — the stored value is never changed, only the displayed representation.
PROC CONTENTS with OUT= saves the dataset descriptor to a dataset, allowing the NAME, LABEL, FORMAT, TYPE, and LENGTH columns to be inspected directly without relying on the output tab.
Both labels and formats can be assigned in a DATA step or updated on an existing dataset using PROC DATASETS without rewriting any observations.
The ATTRIB statement assigns label, format, informat, and length for a variable in a single clause — useful for keeping variable definitions organised in one place.

*Copyright @ www.mycsg.in;

Introduction to SAS - Part 06: Variable Labels, Dataset Labels, and Display Formats

Overview

Variable Labels

SAS Log

Dataset View

Confirming Labels with PROC CONTENTS OUT=

SAS Log

Dataset View

Dataset Labels

SAS Log

Dataset View

Assigning Labels with PROC DATASETS

SAS Log

Dataset View

Display Formats

SAS Log

Dataset View

Assigning Formats with PROC DATASETS

SAS Log

Dataset View

The ATTRIB Statement

SAS Log

Dataset View

Key Points