Missing values

Missing Values and how to work with them

Stata represents missing values with a "." (period) in its results window. If the file has special missing values (available starting with Version 8), Stata represents them as ".a", ".b", ... , ".z". However, missing values are stored as values larger than the largest allowable number for the data type. For example, a variable stored as data type byte can take the following values:

You see:	Stored as:	Treated as:
   1		   1		   1
   2		   2		   2
  ...		  ...		  ...
  100		  100		  100
   .		  101		 missing
   .a		  102		 missing
   .b		  103		 missing
  ...		  ...		  ...
   .z		  127		 missing

Most commands ignore missing values by default. Some commands, such as tabulate, have an option to display missing if you want to see how many missing observations there are. Other commands, however, may use missing values in a way that will surprise you. For example, the replace command does not ignore missing values. Here is a simple example to demonstrate how replace and recode handle missing values differently. In this example, we have three variables with the values of 1, 2, and missing. We want to change all values of 2 to 1.

clear
input a b c
1 1 1
2 2 2
. . .
end
list
replace a=1 if a>1
replace b=1 if b>1 & b<.
recode c 1/max=1
list

The first replace command changes every value that's greater than 1 to 1. This command does not ignore missing values, so both 2 and missing are changed to 1. This probably is not what we would normally want to do, since missing values should remain missing.

The second replace command changes all values greater than 1 but not missing to 1. In this case values of 2 are changed and missing values are not changed, which is our intention.

The recode command automatically ignores missing values, so we don't have to think about it. The results are the same as the second replace command.