Using Informats to Encode and Reorder Character Values

Where the values in a character variable represent category labels, it is sometimes helpful to encode this data as a formatted numeric variable.

Consider the smoking_status variable in the Heart example data set. This variable is a character variable. The values are shown in a frequency table. The values are shown in alphabetic order, but they form an ordinal scale that is most easily understand in “Light” to “Very Heavy” order, or vice versa.

proc freq data=sashelp.heart;
  tables smoking_status / nocum;
run;

                            The FREQ Procedure

                              Smoking Status
 
                Smoking_Status       Frequency     Percent
                ------------------------------------------
                Heavy (16-25)            1046       20.22 
                Light (1-5)               579       11.19 
                Moderate (6-15)           576       11.13 
                Non-smoker               2501       48.35 
                Very Heavy (> 25)         471        9.10 

                          Frequency Missing = 36

It will be easiest to have these categories show up in order in our output if the values are encoded numerically, 1 to 5 (or 0 to 4).

Recoding Numerically

We could recode these data numerically in either of two ways. You are probably already familiar with the use of IF/THEN-ELSE in DATA steps, and that’s a perfectly reasonable approach to this task.

Another approach is to use an informat. The use of SAS informats is discussed elsewhere. Just as you can create your own formats, you can create your own informats. In fact, you use the same PROC FORMAT for both - you probably should use a format with the resulting numeric data, anyway. The difference is that an informat is created with an INVALUE statement.

proc format;
    invalue smok "Non-smoker" = 1
        "Light (1-5)"         = 2
        "Moderate (6-15)"     = 3
        "Heavy (16-25)"       = 4
        "Very Heavy (> 25)"   = 5;
    value smok 1 = "Non-smoker"
        2 = "Light (1-5)"
        3 = "Moderate (6-15)"
        4 = "Heavy (16-25)"
        5 = "Very Heavy (> 25)";
run;

If you create pairs of format/informats, you can give both the same name (just like SAS does). SAS will know which to use based on the context in which it is called.

Informats can only be used in DATA steps, with INPUT functions or INPUT statements. In this DATA step, the character values in smoking_status are converted to numeric values in smoking_rate, and then these numeric values are formatted to show us what the numeric categories mean.

data heart;
    set sashelp.heart;
    smoking_rate = input(smoking_status, smok.);
    format smoking_rate smok.;
run;

The key conversion step here is accomplished with the INPUT function, INPUT(variable, informat.)

ods select variables;
proc contents data=heart(keep=smoking_status smoking_rate);
run;

                          The CONTENTS Procedure

                Alphabetic List of Variables and Attributes
 
       #    Variable          Type    Len    Format    Label

      17    Smoking_Status    Char     17              Smoking Status
      18    smoking_rate      Num       8    SMOK.                   

proc freq data=heart;
  tables smoking_rate / nocum;
run;

                            The FREQ Procedure

                     smoking_rate    Frequency     Percent
                ------------------------------------------
                Non-smoker               2501       48.35 
                Light (1-5)               579       11.19 
                Moderate (6-15)           576       11.13 
                Heavy (16-25)            1046       20.22 
                Very Heavy (> 25)         471        9.10 

                          Frequency Missing = 36

Last Revised: 9/26/2024