Summary Variable Definition Commands

These commands define summary variables that are computed for each (non-empty) cell in a DATAB table. There are five basic DATAB summary variable types: counts; person-year sums; sums; means; and sums-of-squared deviations. In event-time tables, there is a special type of sum variable computed as the sum of a binary indicator which has the value 1 if an event of interest (for example, death from or diagnosis with a specific disease), occurs in a cell. This type of summary variable is an event sum or case count variable. Counts and, for person-year tables, person-year sums are created automatically as the table is defined. The default summary variables in a simple table are %CELLNO and %COUNT. The variable %CELLNO is used by DATAB for data management, the user should not refer to this variable in transformations. The variable %COUNT contains a count of the number of records in each cell. For person-year tables there are three default summary variables: %CELLNO, AT_RISK, and PYR. The variable AT_RISK contains the count of the total number of records ever-at-risk in the cell and PYR is the person-year total for the cell.

In event-count tables, the summary variables associated with a cell are updated for each record categorized into the cell. There are two updating procedures for event-time (person-year) tables: ( ) updates occur only for the cell in which the event occurs, that is, the cell which contains the exit date (exit updating); or () updates occur for each cell in which a record is at risk (risk set updating). The primary example of exit updating is the computation of event sums. Person-year totals or cell-specific exposure means (for all people ever-at-risk) are examples of risk set updating. The XSUM and EVENT commands are used to define exit updated sum variables and the XMEAN command is used to define exit updated means. The RSUM and RMEAN commands can be used to define risk-updated sums and means. In event-count tables, exit updating and risk set updating are equivalent. Sum-of-squared deviation summaries are requested using the SSD options in the XMEAN and RMEAN commands.

The WEIGHT subcommand specifies the weights to be used in computing sums, means, and sum-of-squared deviations. In event-count tables, these summary variables are unweighted unless otherwise specified. By default, in event-time tables, means are updated using person-year weights. In this case, the user can request an unweighted mean by choosing the default count variable AT_RISK as the weight variable.

For event-time tables, the FCOUNT command requests a special count summary variable which contains a count of the number of people first-at-risk in the cell.

The remainder of this section contains detailed information on the Syntax of the DATAB summary variable definition commands.