SAS Programming Structure: DATA and PROC Steps

SAS (Statistical Analysis System) is a powerful software suite widely used in biostatistics for data management, analysis, and reporting. At its core, SAS programs are organized into two main types of steps: DATA steps and PROC steps. These alternate in a program to create datasets, manipulate data, and perform analyses. SAS executes the program sequentially, processing one step at a time.

I'll break this down step by step, with explanations and simple examples. In biostatistics, you'll often use DATA steps for cleaning and preparing datasets (e.g., patient records), and PROC steps for statistical procedures like t-tests or survival analysis.

1. Overview of SAS Program Flow

• A SAS program is a sequence of statements ending with a semicolon (;).


• It starts with optional global statements (e.g., for setting options like page size).


• The bulk consists of DATA steps (for creating or modifying datasets) and PROC steps (for procedures like summaries or regressions).


• Steps are separated by RUN; (or sometimes other terminators like QUIT; for interactive PROCs).


• SAS datasets are stored in "libraries" (e.g., WORK for temporary data).

2. DATA Step: Creating and Manipulating Data

• Purpose:


• Read raw data (from files, keyboards, or existing SAS datasets).


• Perform calculations, conditional logic, or data cleaning.


• Output new datasets.


• Structure:

DATA dataset-name;  /* Names the output dataset */
    /* Input data here (e.g., cards for inline data, INFILE for external files) */
    /* Programming statements: assignments, IF-THEN, DO loops, etc. */
  RUN;  /* Ends the step and executes it */

• How it works: SAS reads an observation, executes the code for it, and writes it to the output dataset. It repeats until all input is processed.


• Biostatistics Example: Suppose you have patient data with age and blood pressure, and you want to create a new variable for hypertension (BP > 140).

DATA patient_data;
    INPUT patient_id age bp;  /* Reads inline data; use INFILE for external files */
    hypertension = (bp > 140);  /* Creates a new logical variable */
    IF age > 65 THEN elderly = 'Yes'; ELSE elderly = 'No';  /* Conditional logic */
  DATALINES;  /* Inline data follows */
  001 45 130
  002 70 150
  003 50 135
  ;  /* Ends data lines */
  RUN;

• Output: A new dataset patientdata with columns: patientid, age, bp, hypertension, elderly.


• In biostats, this is useful for deriving variables like BMI from height/weight or flagging outliers.


• Tips:


• Use SET to read from an existing dataset (e.g., SET old_data;).


• Common functions: MEAN(), SUM(), SUBSTR() for strings.


• For merging datasets: Use MERGE within a DATA step (e.g., for joining patient demographics with lab results).

3. PROC Step: Procedures for Analysis and Reporting

• Purpose:


• Summarize, analyze, or visualize data.


• Generate reports, tables, or models.


• Does not modify the input dataset (it's read-only).


• Structure:

PROC proc-name DATA=dataset-name OPTIONS;  /* Specifies the procedure and input data */
    /* Substatements: e.g., VAR for variables, BY for grouping */
  RUN;  /* Or QUIT; for some PROCs */

• How it works: SAS reads the specified dataset and applies the procedure's logic, producing output (e.g., tables, plots) to the results viewer or files.


• Biostatistics Example: Using the patient_data from above, compute means and perform a t-test on blood pressure by elderly status with PROC MEANS and PROC TTEST.

PROC MEANS DATA=patient_data;  /* Basic summary statistics */
    VAR age bp;  /* Variables to analyze */
    CLASS elderly;  /* Group by this variable */
  RUN;
  PROC TTEST DATA=patient_data;  /* T-test example */
    VAR bp;  /* Response variable */
    CLASS elderly;  /* Grouping variable */
  RUN;

• Output (from PROC MEANS): Table with means, std devs for age and BP, split by elderly status.


• Output (from TTEST): T-statistic, p-value comparing BP between elderly and non-elderly groups—key for hypothesis testing in clinical trials.


• In biostats, common PROCs include:


• PROC FREQ: For categorical data (e.g., chi-square tests on treatment outcomes).


• PROC LOGISTIC: For logistic regression (e.g., modeling disease risk).


• PROC GLM: For ANOVA or linear models (e.g., comparing drug effects).


• Tips:


• Always specify DATA=library.dataset (defaults to last created DATA step).


• Use BY for grouped analyses (requires sorting with PROC SORT first).


• Options like NOPRINT suppress output; ODS for custom reporting (e.g., to PDF).

4. Combining DATA and PROC Steps in a Program

/* Step 1: Create and clean data */
DATA trial_data;
  SET raw_trial;  /* Assume raw_trial exists */
  bmi = weight / (height_cm / 100)**2;  /* Calculate BMI */
  IF response = 'Yes' THEN success = 1; ELSE success = 0;  /* Binary outcome */
RUN;
/* Step 2: Sort for grouping */
PROC SORT DATA=trial_data;
  BY treatment_group;
RUN;
/* Step 3: Analyze */
PROC FREQ DATA=trial_data;
  TABLES treatment_group * success / CHISQ;  /* Contingency table and chi-square */
RUN;
PROC MEANS DATA=trial_data;
  VAR bmi;
  BY treatment_group;
RUN;

• This reads raw data, adds variables, sorts, then runs frequency and means analyses—common in preparing for regulatory submissions.

5. Best Practices and Learning Tips

• Error Handling: SAS logs errors in the log window; check for "NOTE" vs. "ERROR" messages.


• Libraries: Use LIBNAME mylib 'path'; for permanent storage (e.g., saving cleaned biostats datasets).


• Debugging: Add PUT statements in DATA steps to print variable values (e.g., PUT age= bp=;).


• Resources for Biostats:


• SAS Documentation: Search "SAS biostatistics" on support.sas.com for PROC examples.


• Books: "The Little SAS Book" for basics; "SAS for Clinical Trial Data Management" for biostats-specific.


• Practice: Download free SAS University Edition or use SAS OnDemand for Academics. Try sample datasets from clinical trials (e.g., from FDA sites).


• Common Pitfalls: Forgetting semicolons, mismatched DATA/PROC datasets, or not sorting before BY-group processing.