SAS Formats and Informats

An informat is a specification for how raw data should be read. A format is a layout specification for how a variable should be printed or displayed.

SAS contains many internal formats and informats, or user defined formats and informats can be constructed using PROC FORMAT. To see a list of all internal formats and informats, type 'help format' in the command line and then click on 'SAS Formats and Informats' in the resulting window.

The following code shows how SAS informats can be used:

input  @1 pnr       10.
       @11 sex       1.
       @12 surname  $15
       @27 diadate  yymmdd6.

Formats are much more widely used in our department then informats.

The following code shows how a user-defined format can be created.

proc format;
value sex

The code above only creates the format, it does not associate it with any variable. Formats can be associated with variables in either data steps or proc steps.

Assume we create a SAS data set in a data step, and include the following line in the data step.

format bmi 6.2 sex sex.;

We have applied the SAS system format 6.2 to the variable bmi and the user-defined format sex to the variable sex. Whenever we display information from this data set (using PROC FREQ or PROC PRINT, for example) bmi will always be rounded to 2 decimal places and sex will be displayed according to the given format. Note that applying a format does not affect the value of the variable. BMI will still be stored using many significant digits and sex will be stored as an integer between 0 and 9.

We can also specify formats in individual proc steps, which override the default format (if a default format exists). We could, for example, use the following code, which would produce a listing of the actual values of the variable sex.

proc freq;
tables sex;
format sex 2.;

If we did not assign formats to the variables bmi or sex in a data step, we can assign them is a proc step, in which case the formats are only used within that proc step. For example:

proc freq;
tables sex bmi;
format bmi 6.2 sex sex.;

Any calculations made using a variable in a data step will be based on the raw data (i.e. the format is ignored). When fitting statistical models, however, the model is fitted to the formatted value.