Variables, Types, and Missing Values

SAS has just two data types:

• Numeric — stored as 8-byte floating-point numbers. Dates and times are numeric too.
• Character — fixed-length text, padded with blanks to the variable's length (1 to 32,767 bytes).

A variable's type and length are fixed the first time SAS sees it in a step.

Missing values

A missing numeric value is written . and a missing character value is a blank. Arithmetic with a missing value gives missing — but SAS functions like SUM and MEAN skip missing values:

data totals;
   input name $ test1 test2;
   total = test1 + test2;
   total2 = sum(test1, test2);
   datalines;
Ada 90 85
Alan . 70
;
run;

proc print data=totals;
run;

Obs    name    test1    test2    total    total2

  1    Ada        90       85      175       175
  2    Alan        .       70        .        70

And the log warns you:

NOTE: Missing values were generated as a result of performing an operation on missing values.

Neither result is "right" — decide whether a missing test should count as zero (SUM) or make the total unknown (+).

Missing is smaller than every number

In comparisons, a missing numeric value is lower than any number, including negative ones. That catches everyone once:

data check;
   input name $ score;
   length status $ 4;
   if score < 50 then status = 'fail';
   else status = 'pass';
   datalines;
Ada 72
Alan .
Grace 41
;
run;

proc print data=check;
run;

Obs    name     score    status

  1    Ada         72     pass
  2    Alan         .     fail
  3    Grace       41     fail

Alan has no score, yet he "failed". Test for missing values first:

if missing(score) then status = 'n/a';
else if score < 50 then status = 'fail';
else status = 'pass';

MISSING() works for both numeric and character variables.

Automatic type conversion

If you use a character value where a number is expected (or the reverse), SAS converts it and writes a note:

data _null_;
   x = '42' + 1;
   put x=;
run;

Log (abridged):

NOTE: Character values have been converted to numeric values at the places given by: (Line):(Column).
x=43

It works, but treat that note as a bug to fix: convert explicitly with INPUT() and PUT() (covered in Level 2), so your intent is clear.

Practice

Read three products with a price and a discount, where one discount is missing. Compute the final price so that a missing discount counts as no discount.

data prices;
   input product $ price discount;
   final = price - sum(discount, 0);
   datalines;
Lamp 40 5
Desk 150 .
Chair 60 10
;
run;

proc print data=prices;
run;

Obs    product    price    discount    final

  1    Lamp          40           5       35
  2    Desk         150           .      150
  3    Chair         60          10       50