User guide > Weighting > Respondent weighting

Respondent weighting occurs when a respondent's answers are weighted according to their sampling characteristics (for example age and class). Typically each respondent will use one respondent weight for a whole set of tables, and this weight is pointed to by SELECT WR.

Different variables, and the number 1 (meaning no respondent weight) can be assigned using the SELECT WR command. This command is non-executable, and specifies that for all T (tabulate) commands following down to the next SELECT WR command, the variable specified contains the respondent weight. The variable named can be redefined as often as necessary.

If a respondent weight is specified for a table, two tables are held on disk for that table, one weighted and one unweighted. Nearly all operations specified for a table affect both the weighted and unweighted table, but it is the weighted one that is usually printed. Unweighted figures can be shown on the table if formats UNR, UNC, PUC or RWU are used.

Note that use of the SELECT WR increases the amount of disk storage required for the .ITF.

If weighting will bring the grand total for the table to over a million, you are advised to use format DBL.

Examples:

select wr $weight,

select wr 1,