You've learned how to create a SAS data set from raw data that is stored in an external file. You've also learned how to subset observations and how to assign values to variables.

This chapter shows you additional techniques for creating and managing variables. In this chapter, you learn how to create sum variables, assign variable values conditionally, select variables, and assign permanent labels and formats to variables.

In this chapter, you learn how to

It is often useful to create a variable that accumulates the values of another variable. Suppose you want to create the data set Clinic.Stress and to add a new variable, SumSec, to accumulate the total number of elapsed seconds in treadmill stress tests.

To add the result of an expression to an accumulator variable, you can use a sum statement in your DATA step.

General form, sum statement:

   variable+expression;

where

• variable specifies the name of the accumulator variable. This variable must be numeric. The variable is automatically set to 0 before the first observation is read. The variable's value is retained from one DATA step execution to the next.

• expression is any valid SAS expression.

The sum statement adds the result of the expression that is on the right side of the plus sign (+) to the numeric variable that is on the left side of the plus sign. At the top of the DATA step, the value of the numeric variable is not set to missing as it usually is when reading raw data. Instead, the variable retains the new value in the program data vector for use in processing the next observation.

To find the total number of elapsed seconds in treadmill stress tests, you need a variable (in this example, SumSec) whose value begins at 0 and increases by the amount of the total seconds in each observation. To calculate the total number of elapsed seconds in treadmill stress tests, you use the sum statement shown below:

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
            RecHR 35-37 TimeMin 39-40 TimeSec 42-43
            Tolerance $ 45;
   TotalTime=(timemin*60)+timesec;
   SumSec+totaltime;
run;

The value of the variable on the left side of the plus sign (here, SumSec) begins at 0 and increases by the value of TotalTime with each observation.

In a previous example, the sum variable SumSec was initialized to 0 by default before the first observation was read. But what if you want to initialize SumSec to a different number, such as the total seconds from previous treadmill stress tests?

You can use the RETAIN statement to assign an initial value other than the default value of 0 to a variable whose value is assigned by a sum statement.

The RETAIN statement

General form, simple RETAIN statement for initializing sum variables:

   RETAIN variable <initial;-value>;

where

• variable is a variable whose values you want to retain

• initial-value specifies an initial value (numeric or character) for the preceding variable.

• is a compile-time only statement that creates variables if they do not already exist.

• initializes the retained variable to missing before the first execution of the DATA step if you do not supply an initial value.

• has no effect on variables that are read with SET, MERGE, or UPDATE statements. (The SET and MERGE statements are discussed in later chapters.)

Suppose you want to add 5400 seconds (the accumulated total seconds from a previous treadmill stress test) to the variable SumSec in the Clinic.Stress data set when you create the data set. To initialize SumSec with the value 5400, you use the RETAIN statement shown below.

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33

         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
    TotalTime=(timemin*60)+timesec;
    retain SumSec 5400;
    sumsec+totaltime;
run;

Now the value of SumSec begins at 5400 and increases by the value of TotalTime with each observation.

To assign the value Long to the variable TestLength when the value of TotalTime is greater than 800, add the following IF-THEN statement to your DATA step:

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   if totaltime>800 then TestLength='Long';
run;

SAS executes the assignment statement only when the condition (TotalTime>800) is true. If the condition is false, the value of TestLength will be missing.

When writing IF-THEN statements, you can use any of the following comparison operators:

if test<85 and time<=20
   then Status='RETEST';
if region in ('NE','NW','SW')
   then Rate=fee-25;
if target gt 300 or sales ge 50000
   then Bonus=salary*.05;

You can also use these logical operators:

Use the AND operator to execute the THEN statement if both expressions that are linked by AND are true.

if status='OK' and type=3
   then Count+1;
if (age^=agecheck | time^=3)
   & error=1 then Test=1;

Use the OR operator to execute the THEN statement if either expression that is linked by OR is true.

if (age^=agecheck | time^=3)
    & error=1 then Test=1;
if status='S' or cond='E'
   then Control='Stop';

Use the NOT operator with other operators to reverse the logic of a comparison.

if not(loghours<7500)
    then Schedule='Quarterly';
if region not in ('NE','SE')
    then Bonus=200;

Character values must be specified in the same case in which they appear in the data set and must be enclosed in quotation marks.

if status= 'OK' and type=3
   then Count+1;
if status= 'S' or cond= 'E'
   then Control= 'Stop';
if not(loghours<7500)
   then Schedule= 'Quarterly';
if region not in ('NE','SE')
   then Bonus=200;

Logical comparisons that are enclosed in parentheses are evaluated as true or false before they are compared to other expressions. In the example below, the OR comparison in parentheses is evaluated before the first expression and the AND operator are evaluated.

0 = False

. = False

1 = True

As a result, you need to be careful when using the OR operator with a series of comparisons. Remember that only one comparison in a series of OR comparisons must be true to make a condition true, and any nonzero, nonmissing constant is always evaluated as true. Therefore, the following subsetting IF statement is always true:

if x=1 or 2;

SAS first evaluates x=1, and the result can be either true or false; however, since the 2 is evaluated as nonzero and nonmissing (true), the entire expression is true. In this statement, however, the condition is not necessarily true because either comparison can evaluate as true or false:

if x=1 or x=2;

Now suppose you want to assign a value to TestLength based on the other possible values of TotalTime. One way to do this is to add IF-THEN statements for the other two conditions, as shown below.

if totaltime>800 then TestLength='Long';
if 750<=totaltime<=800 then TestLength='Normal';
if totaltime<750 then TestLength='Short';

However, when the DATA step executes, each IF statement is evaluated in order, even if the first condition is true. This wastes system resources and slows the processing of your program.

Instead of using a series of IF-THEN statements, you can use the ELSE statement to specify an alternative action to be performed when the condition in an IF-THEN statement is false. As shown below, you can write multiple ELSE statements to specify a series of mutually exclusive conditions.

if totaltime>800 then TestLength='Long';
else if 750<=totaltime<=800 then TestLength='Normal';
else if totaltime<750 then TestLength='Short';

The ELSE statement must immediately follow the IF-THEN statement in your program. An ELSE statement executes only if the previous IF-THEN/ELSE statement is false.

So, to assign a value to TestLength when the condition in your IF-THEN statement is false, you can add the ELSE statement to your DATA step, as shown below:

data clinic.stress;
     infile tests;

     input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
           RecHR 35-37 TimeMin 39-40 TimeSec 42-43
           Tolerance $ 45;
     TotalTime=(timemin*60)+timesec;
     retain SumSec 5400;
     sumsec+totaltime;
     if totaltime>800 then TestLeng h='Long';
     else if 750<=totaltime<=800 then TestLength='Normal'; 
     else if totaltime<750 then TestLength='Short';
run;

For greater efficiency, construct your IF-THEN/ELSE statements with conditions of decreasing probability.

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   if totaltime>800 then TestLength='Long';
   else if 750<=totaltime<=800 then TestLength='Normal';
   else put 'NOTE: Check this Length: 'totaltime=;
run;

length Type $ 8;
length Address1 Address2 Address3    $200;
length  FirstName $12 LastName  $16;

Within your program, you include a LENGTH statement to assign a length to accommodate the longest value of the variable TestLength. The longest value is Normal, which has six characters. Because TestLength is a character variable, you must follow the variable name with a dollar sign ($).

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   length TestLength $ 6;
   if totaltime>800 then testlength='Long';
   else if 750<=totaltime<=800 then testlength='Normal';
   else if totaltime<750 then TestLength='Short';
run;

Now that you have added the LENGTH statement to your program, the values of TestLength are no longer truncated.

So far in this chapter, you've learned to use IF-THEN statements to execute assignment statements conditionally. But you can specify any executable SAS statement in an IF-THEN statement. For example, you can use an IF-THEN statement with a DELETE statement to determine which observations to omit as you read data.

The IF-THEN and DELETE statements below omit any observations whose values for RestHR are lower than 70.

data clinic.stress;
   infile tests;
    input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
          RecHR 35-37 TimeMin 39-40 TimeSec 42-43
          Tolerance $ 45;
    if resthr<70 then delete;
    TotalTime=(timemin*60)+timesec;
    retain SumSec 5400;
    sumsec+totaltime;
    length TestLength $ 6;
    if totaltime>800 then testlength='Long';
    else if 750<=totaltime<=800 then testlength='Normal';
    else if totaltime<750 then TestLength='Short';
run;

Sometimes you may need to read and process fields that you don't want to keep in your data set. In this case, you can use the DROP= and KEEP= data set options to specify the variables that you want to drop or keep.

Use the KEEP= option instead of the DROP= option if more variables are dropped than kept. You specify data set options in parentheses after a SAS data set name.

Suppose you are interested in keeping only the new variable TotalTime and not the original variables TimeMin and TimeSec. You can drop TimeMin and TimeSec when you create the Stress data set.

data clinic.stress(drop=timemin timesec);
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   if tolerance='D';
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   length TestLength $ 6;
   if totaltime>800 then testlength='Long';
   else if 750<=totaltime<=800 then testlength='Normal';
   else if totaltime<750 then TestLength='Short';
run;

Another way to exclude variables from your data set is to use the DROP statement or the KEEP statement. Like the DROP= and KEEP= data set options, these statements drop or keep variables. However, the DROP statement differs from the DROP= data set option in the following ways:

The KEEP statement is similar to the DROP statement, except that the KEEP statement specifies a list of variables to write to output data sets. Use the KEEP statement instead of the DROP statement if the number of variables to keep is significantly smaller than the number to drop.

The two programs below produce the same results. The first example uses the DROP= data set option; the second example uses the DROP statement.

data clinic.stress(drop=timemin timesec);
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   if tolerance='D';
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   length TestLength $ 6;
   if totaltime>800 then testlength='Long';
   else if 750<=totaltime<=800 then testlength='Normal';
   else if totaltime<750 then TestLength='Short';
run;

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   if tolerance='D';
   drop timemin timesec;
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   length TestLength $ 6;
   if totaltime>800 then testlength='Long';
   else if 750<=totaltime<=800 then testlength='Normal';
   else if totaltime<750 then TestLength='Short';
run;

To specify the label Cumulative Total Seconds (+5,400) and the format COMMA6. for the variable SumSec, you can submit the following program:

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;

   if resthr<70 then delete;
   if tolerance='D';
   drop timemin timesec;
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   length TestLength $ 6;
   if totaltime>800 then testlength='Long';
   else if 750<=totaltime<=800 then testlength='Normal';
   else if totaltime<750 then TestLength='Short';
   label sumsec='Cumulative Total Seconds (+5,400)'; 
   format sumsec comma6.;
run;

You're done! When you print the new data set, SumSec is labeled and formatted as specified. (Don't forget the LABEL option in the PROC PRINT statement.)

proc print data=clinic.stress label;
 run;

The following code is a simple example of a SELECT group. Notice that the variable a is specified in the SELECT statement, and various values to compare to a are specified in the WHEN statements. When the value of the variable a is

select (a);
   when (1) x=x*10;
   when (3,4,5) x=x*100;
   otherwise;
end

Now let's look at a SELECT group in context. In the DATA step below, the SELECT group assign values to the variable Group based on values of the variable JobCode.

Most of the assignments are one-to-one correspondences, but ticket agents (the JobCode values TA1, TA2, and TA3) are grouped together, as are values in the Other category.

data emps(keep=salary group);
   set sasuser.payrollmaster;
   length Group $ 20;
   select(jobcode);
      when ("FA1") group="Flight Attendant I";
      when ("FA2") group="Flight Attendant II";
      when ("FA3") group="Flight Attendant III";
      when ("ME1") group="Mechanic I";
      when ("ME2") group="Mechanic II";
      when ("ME3") group="Mechanic III";
      when ("NA1") group="Navigator I";
      when ("NA2") group="Navigator II";
      when ("NA3") group="Navigator III";
      when ("NA1") group="Navigator I";
      when ("NA2") group="Navigator II";
      when ("NA3") group="Navigator III";
      when ("PT1") group="Pilot I";
      when ("PT2") group="Pilot II";
      when ("PT3") group="Pilot III";
      when ("TA1","TA2","TA3") group="Ticket Agents";
      otherwise group="Other";
    end;
run;

Notice that in this case the SELECT statement does contain a select-expression. You are checking values of a single variable, so here using select(jobcode) and only the JobCode value in each WHEN statement is more concise than eliminating the select-expression and repeating the variable in each when-expression, as in when(jobcode="FA1").

when ("fa1") group="Flight Attendant I";

In this case, given the SELECT group above, Group would be assigned the value Other.

As you saw in the general form for SELECT groups, you can optionally specify a select-expression in the SELECT statement. The way SAS evaluates a when-expression depends on whether you specify a select-expression.

If you do specify a select-expression in the SELECT statement, SAS compares the value of the select-expression with the value of each when-expression. That is, SAS evaluates the select-expression and when-expression, compares the two for equality, and returns a value of true or false.

In the following SELECT group, SAS determines the value of toy and compares it to values in each WHEN statement in turn. If a WHEN statement is true compared to the toy value, SAS assigns the related price and continues processing the rest of the DATA step. If none of the comparisons is true, SAS executes the OTHERWISE statement and writes a debugging message to the SAS log.

select (toy);
   when ("Bear") price=35.00;
   when ("Violin") price=139.00;
   when ("Top","Whistle","Duck") price=7.99;
   otherwise put "Check unknown toy: " toy=;
end;

If you don't specify a select-expression, SAS evaluates each when-expression to produce a result of true or false.

If more than one WHEN statement has a true when-expression, only the first WHEN statement is used; once a when-expression is true, no other when-expressions are evaluated.

In the example below, the SELECT statement does not specify a select-expression. The WHEN statements are evaluated in order, and only one is used. For example, if the value of toy is Bear and the value of month is FEB, only the second WHEN statement is used, even though the condition in the third WHEN statement is also met. In this case, the variable price is assigned the value 25.00:

select;
   when (toy="Bear" and month in ('OCT', 'NOV', 'DEC')) price=45.00;
   when (toy="Bear" and month in ('JAN', 'FEB')) price=25.00;
   when (toy="Bear") price=35.00;
   otherwise;
end;

In this simple DO group, the statements between DO and END are performed only when TotalTime is greater than 800. If TotalTime is less than or equal to 800, statements in the DO group do not execute, and the program continues with the assignment statement that follows the appropriate ELSE statement.

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   length TestLength $ 6 Message $ 20;
   if totaltime>800 then
      do; 
         testlength='Long'; 
         message='Run blood panel'; 
      end;
    else if 750<=totaltime<=800 then testlength='Normal';
    else if totaltime<750 then TestLength='Short';
run;

In the SELECT group below, the statements between DO and END are performed only when the value of Payclass is hourly. Notice that an IF-THEN statement appears in the DO group; the PUT statement executes only when Hours is greater than 40. The second END statement in the program closes the SELECT group.

In the SELECT group below, the statements between DO and END are performed only when the value of Payclass is hourly. Notice that an IF-THEN statement appears in the DO group; the PUT statement executes only when Hours is greater than 40. The second END statement in the program closes the SELECT group.

data payroll;
   set salaries;
   select(payclass);
   when ('monthly') amt=salary;
   when ('hourly')
      do; 
         amt=hrlywage*min(hrs,40); 
         if hrs>40 then put 'CHECK TIMECARD'; 
      end;
   otherwise put 'PROBLEM OBSERVATION';
   end;
run;

You can nest DO groups to any level, just like you nest IF-THEN/ELSE statements. (The memory capabilities of your system may limit the number of nested DO statements you can use. For details, see the SAS documentation about how many levels of nested DO statements your system's memory can support.)

The following is an example of nested DO groups:

do;
    statements;
       do;
                statements ;
                   do;
                           statements;
                   end;
       end;
end;

It is good practice to indent the statements in DO groups, as shown in the preceding statements, so that their position indicates the levels of nesting.

You can learn about these forms of the DO statement in Chapter 15, "Generating Data with DO Loops," on page 473.

LIBNAME libref 'SAS-data-library';

FILENAME fileref 'filename';

DATA SAS-data-set(DROP=variable(s)|KEEP=variable(s));

        INFILE file-specification <OBS=n>;

        INPUT variable < $> startcol-endcol...;

        DROP variable(s);

        KEEP variable(s);

        RETAIN variable initial-value;

        variable+expression;

        LENGTH variable(s) <$> length;

        IF expressionTHEN statement;

        ELSE statement;

        IF expressionTHEN DELETE;

        LABEL variable1='label1' variable2='label2' ...;

        FORMAT variable(s) format-name;

        SELECT <(select-expression)>;

            WHEN-1 (when-expression-1 <..., when-expression-n>) statement;

            WHEN-n (when-expression-1 <..., when-expression-n) statement; >;

            <OTHERWISE statement;>

        END;

RUN;

PROC PRINT DATA=SAS-data setLABEL;

RUN;

data clinic.stress;
   infile tests;
   input ID $ 1-4 Name $ 6-25 RestHR 27-29 MaxHR 31-33
         RecHR 35-37 TimeMin 39-40 TimeSec 42-43
         Tolerance $ 45;
   if tolerance='D'and resthr ge 70 then delete;
   drop timemin timesec;
   TotalTime=(timemin*60)+timesec;
   retain SumSec 5400;
   sumsec+totaltime;
   length TestLength $ 6;
   if totaltime>800 then testlength='Long';
   else if 750<=totaltime<=800 then testlength='Normal';
   else if totaltime<750 then TestLength='Short';
   label sumsec='Cumulative Total Seconds (+5,400)';
   format sumsec comma6.;
run;