SAS Data Step

In SAS, data is manipulated in data steps.

---

Data Model

The data model for SAS is:

• Read a row of data
• Process statements sequentially

• On run, write data into the output table

data LIBREF.TABLE;
  STATEMENTS;
run;

---

Tables

The table named in a data step is always the output table. By default it is also the input table and data is manipulated in place.

To set a different table as the input, use a set statement.

data LIBREF.OUTPUT;
  set LIBREF.INPUT;
  STATEMENTS;
run;

---

Data Input

See here.

---

Transforming Data

Creating Variables

data LIBREF.OUTPUT;
  set LIBREF.INPUT;

  /* declare numeric variables directly with numeric literals */
  foo = 1;

  /* declare string variables by specifying the length first, then with string literals */
  length hello $5 world $5;
  hello = 'Hello';
  world = 'World';

  /* the length of a string variable cannot be changed again within the same data step */
  length greeting $12;
  greeting = CAT(hello, world);        /* 'Hello World'  */
  greeting = CATS(hello, world);       /* 'HelloWorld'   */
  greeting = CATX(', ', hello, world); /* 'Hello, World' */
  greeting = hello||world;             /* 'HelloWorld'   */
run;

Any valid expression can be used to create a new variable.

Coercing Data Types

data LIBREF.NEWTABLE;
  set LIBREF.OLDTABLE;

  /* coerce string to numeric */
  numeric_zip_code = input(string_zip_code, 5.);

  /* left-align text */
  length TEXT_VAR $999.;
  TEXT_VAR = put(TEXT_VAR, $999. -L);

  /* parse timestamps like '01Jan1999' */
  format timestamp DATE9.;
  month = month(timestamp);

  /* round numerics to integers */
  format test_score 3.;
run;

The input function generates a variable _ERROR_ by default, flagging cases that could not be formatted. To suppress this variable's creation, use input(ALNUM_ID, ?? 8.).

---

Joining Data

See here.

---

Filtering Data

Subset Cases

To subset the cases of a data table, use where statements.

data LIBREF.NEWTABLE;
  set LIBREF.OLDTABLE;
  where EXPR;
run;

Note that where can only be used with variables that exist before the data step.

if statements can be used similarly and also bypass that last restriction. The downsides of this approach are that the operation is more computationally expensive, and there isn't a clear analogy to the where statements on procedures.

data LIBREF.NEWTABLE;
  set LIBREF.OLDTABLE;
  NEWVAR=1;
  if NEWVAR=1;
run;

Alternatively, toggle the output data table with if statements.

data LIBREF.NEWTABLE1;
  set LIBREF.OLDTABLE;
  if EXPR then output LIBREF.NEWTABLE1; else output LIBREF.NEWTABLE2;
run;

Subset Variables

To subset the variables of a data table, use keep and drop statements.

data LIBREF.NEWTABLE;
  set LIBREF.OLDTABLE;
  keep VARLIST1;
  drop VARLIST2;
run;

De-duplication

This is actually best done using the SORT procedure. See here for details.

---

CategoryRicottone