3.1 Reading SAS Data Sets with the SET Statement

3.1 Reading SAS Data Sets with the SET Statement

Quick Answer: The SET statement reads observations sequentially from one or more existing SAS data sets into the Program Data Vector (PDV). Unlike raw data read with INPUT, variables read via SET automatically retain their values across DATA step iterations until replaced by subsequent observations.

Understanding how SAS reads existing SAS data sets using the SET statement is a cornerstone of SAS Base programming and a primary focus of the SAS Certified Specialist exam (A00-231). The SET statement operates inside a DATA step to read data sequentially from an input data set, copy variable descriptor information, and populate the Program Data Vector (PDV).

---

1. Compilation Phase vs. Execution Phase Mechanics

SAS processes a DATA step containing a SET statement in two distinct phases: Compilation and Execution.

data work.employee_copy;
    set sashelp.class;
run;

Compilation Phase

During compilation, SAS scans the code to establish data structures before reading any observations:

1. Descriptor Creation: SAS opens the header descriptor of the input dataset (sashelp.class). It copies the variable attributes—including variable names, types (character or numeric), lengths, formats, informats, and labels—to the descriptor portion of the new output dataset (work.employee_copy).
2. PDV Construction: SAS creates the Program Data Vector (PDV) layout based on the discovered variable attributes.
3. Automatic Variables: SAS adds two hidden automatic variables to the PDV:
   • _N_: A numeric iteration counter initialized to 1.
   • _ERROR_: A numeric binary flag initialized to 0, which signals data or processing errors during execution.

Execution Phase

During execution, SAS iterates through the input dataset line by line:

1. PDV Initialization: At the beginning of the FIRST iteration (_N_=1), all variables in the PDV are set to missing (. for numeric, blank " " for character).
2. SET Execution: The SET statement reads the first observation from sashelp.class directly into the matching PDV variable slots.
3. Program Logic: SAS executes any subsequent programming statements in the DATA step.
4. Implicit OUTPUT: Upon reaching the run; statement (or implicit RETURN), SAS executes an implicit OUTPUT statement, copying the contents of the PDV as a row in the output dataset.
5. Iteration & Retention: SAS returns to the top of the DATA step (RETURN), increments _N_ by 1, and executes SET to read the next observation.

---

2. Variable Retention Behavior in the PDV

A critical exam rule distinguishes variables created by assignment statements from variables read from SAS data sets via SET:

Exam Rule: Variables read from an existing SAS data set using the SET statement are automatically retained across iterations. SAS does NOT re-initialize input dataset variables to missing at the beginning of subsequent DATA step iterations.

In contrast, newly calculated variables created via assignment statements (e.g., temp_calc = age * 10;) are re-initialized to missing at the start of every iteration unless explicitly preserved using a RETAIN statement or a sum statement (+).

data work.calc_example;
    set sashelp.class;
    
    /* Variable created by assignment: reset to missing each iteration */
    temp_calc = age * 10;
    
    /* Variable created by sum statement: automatically retained */
    cumulative_age + age;
run;

---

3. Dataset Options Used with the SET Statement

SAS provides specialized dataset options on the SET statement to control data stream reading and last-observation logic.

The END= Option

The END= option defines a temporary, numeric indicator variable that flags when SAS reaches the final observation of the input dataset.

data work.summary_stat;
    set sashelp.class end=last_rec;
    total_weight + weight;
    
    /* Output only the single summary row when last_rec equals 1 */
    if last_rec = 1 then output;
run;

• last_rec holds a value of 0 for all observations except the final observation, where it equals 1.
• END= variables exist only in the PDV during DATA step execution; they are never written to the output dataset.

Direct Access with POINT= and NOBS=

The POINT= option specifies a numeric variable containing the observation number to read directly (random access). The NOBS= option creates a temporary variable containing the total count of observations in the dataset.

data work.sample_middle;
    /* NOBS assigns total record count to total_obs at compilation */
    set sashelp.class nobs=total_obs;
    
    /* Fetch observation #5 directly */
    pick_obs = 5;
    set sashelp.class point=pick_obs;
    
    output;
    stop; /* MANDATORY: Prevents infinite loop */
run;

Warning / Exam Trap: When using POINT=, SAS does not encounter an End-Of-File (EOF) marker. Without an explicit STOP statement, the DATA step will enter an infinite loop, repeatedly executing the step without terminating!

---

4. Common Exam Pitfalls & Traps

1. Length Mismatches Across Data Steps: If a variable is assigned a length in a preceding statement before SET, SAS retains that initial length. If the input dataset has a longer length, character string truncation occurs without warning.
2. Conditional SET Statements: Placing SET inside an IF-THEN block can cause unexpected execution behavior, as observations are only read when the condition evaluates to true.
3. Misinterpreting _N_: _N_ records the number of times the DATA step has executed, not necessarily the observation number of the output dataset (especially if conditional OUTPUT or subsetting WHERE statements are present).