3.4 Controlling Data Output & Variable Selection (KEEP, DROP, RENAME, OBS=)

3.4 Controlling Data Output & Variable Selection (KEEP, DROP, RENAME, OBS=)

Quick Answer: Use KEEP/DROP statements or options to select variables, RENAME to change variable names, and OBS=/FIRSTOBS= to control row processing. Statement-level options affect all output datasets, while dataset options target specific input or output streams.

Controlling variable selection, variable naming, and observation counts optimizes processing speed, reduces memory usage, and ensures correct data structure. Understanding the precise timing of when KEEP, DROP, RENAME, and OBS= operate in the Program Data Vector (PDV) is vital for the SAS Base Programming exam.

---

1. Variable Selection: Statements vs. Dataset Options

You can exclude or include variables using either Statements (KEEP, DROP) or Dataset Options ((KEEP=...), (DROP=...)).

/* Statement Level */
data work.subset1;
    set sashelp.class;
    keep Name Age Height;
run;

/* Dataset Option Level */
data work.subset2;
    set sashelp.class (keep=Name Age Height);
run;

Critical Difference: Input vs. Output Scope

/* Input Dataset Option Example: Variable NOT in PDV */
data work.calc_fail;
    set sashelp.class (keep=Name Age);
    
    /* ERROR / MISSING: Weight was dropped BEFORE entering PDV */
    BMI = (Weight / (Height**2)) * 703; 
run;

Exam Rule: If a variable is dropped using an input dataset option (SET ds(DROP=Weight)), that variable is never loaded into the PDV. Any subsequent DATA step logic referencing Weight will evaluate to missing or throw a compiler warning!

---

2. Variable Renaming with RENAME

The RENAME statement or RENAME= option changes variable names.

/* Statement level */
data work.renamed1;
    set sashelp.class;
    rename Name=StudentName Age=YearsOld;
run;

/* Input Dataset Option level */
data work.renamed2;
    set sashelp.class (rename=(Name=StudentName Age=YearsOld));
    
    /* MUST use the NEW name in subsequent programming statements! */
    if YearsOld > 13; 
run;

Exam Trap: When RENAME= is used as an input dataset option, the variable name changes immediately as it enters the PDV. All subsequent programming logic inside the DATA step MUST reference the NEW variable name (YearsOld), not the old name (Age).

---

3. Controlling Data Output with Explicit OUTPUT Statements

By default, SAS automatically writes the contents of the PDV to the output dataset at the end of each DATA step iteration (implicit output).

/* Implicit Output */
data work.implicit;
    set sashelp.class;
    /* Implicit OUTPUT occurs at run; */
run;

/* Explicit Output */
data work.boys work.girls;
    set sashelp.class;
    if Sex = 'M' then output work.boys;
    else if Sex = 'F' then output work.girls;
run;

Rules of Explicit OUTPUT Statements

1. Suppression of Default Output: As soon as SAS encounters ANY explicit OUTPUT statement in a DATA step, the default implicit output at the end of the iteration is completely disabled.
2. Multiple Outputs Per Iteration: A single DATA step iteration can execute multiple OUTPUT statements, producing multiple rows in output datasets from a single input observation (e.g., restructuring wide data to long data).

/* Writing Multiple Rows per Iteration */
data work.quarterly_long;
    set work.yearly_wide;
    
    Quarter = 1; Sales = Q1_Sales; output;
    Quarter = 2; Sales = Q2_Sales; output;
    Quarter = 3; Sales = Q3_Sales; output;
    Quarter = 4; Sales = Q4_Sales; output;
run;

---

4. Subsetting Observations with FIRSTOBS= and OBS=

The FIRSTOBS= and OBS= dataset options specify observation processing boundaries.

• FIRSTOBS=n: Specifies the starting observation number (default is 1).
• OBS=m: Specifies the ending observation number.

data work.sample_range;
    set sashelp.class (firstobs=5 obs=15);
run;

• Total observations processed = OBS - FIRSTOBS + 1 = 15 - 5 + 1 = 11 observations.

---

5. Summary Comparison of Output Control Features