Data controls

Data are defined as vectors of differing lengths of data observations on data carriers. Each vector consists of a sequence of cases, and there are two types of case: (1) a real number, representing an observation; (2) undefined. For example that might be the data carrier with data vector (largest case 6) shown in Figure 20.1.

 
Figure 20.1:  A fictitious data vector.

Construct data for a new element

 

Usage

• BD>control  -builddata  switches off the control.

• BD>control  builddata  switches on the control.

• default setting is ON.

The builddata  control is used to determine whether and how data is treated when an element is defined by a BUILD:  command or a PRODUCT:  command. For the PRODUCT:  command there are two possibilities. If the builddata  control is switched on, the program will construct data as appropriate and where possible. Otherwise no data is constructed. Under the BUILD:  command there are four possibilities, depending upon whether or not the control is switched on; and whether or not the element being defined is entirely new, or whether it is intended to overwrite a pre-existent element having the same name.

• The element preexists and the builddata  control is switched on: an attempt is made to construct data from the definition of the RHS in the BUILD:  command. Any previously existent data is overwritten or removed.
• The element preexists and the builddata  control is switched off: any data which had been defined for the pre-existent element becomes attached to the new definition, and there is no attempt made to construct new data.
• The element is entirely new and the builddata  control is switched on: an attempt is made to construct data from the definition of the RHS in the BUILD:  command.
• The element is entirely new and the builddata  control is switched off: no new data is constructed. However, if a data carrier with the same name as the new element already exists, then the data involved become attached to the new element, and the data carrier is removed.

In circumstances where an attempt is made to construct data from the definition part of the BUILD:  command, the maximum possible subset of cases for which there exists data is used. See Figure 6.3 for the complete scheme.  

Automatic data selection

 

Usage

• BD>control  -autoselect  switches off the control.

• BD>control  autoselect  switches on the control.

• default setting is ON.

Many of the commands which use data or create it are subject to the autoselect  control. When the control is switched off, the data selection is whatever has been established by using the SELECT:  command, and the total number of selected cases is accessible as the obs  operand.

When the control is switched on, data are temporarily selected according to the number of observations on a particular key data carrier: typically the first data carrier encountered during a particular command. For the different commands, the autoselection is defined as follows:

• ADJUST:  the first element encountered. (If base notation is used for elements possessing data, elements are taken in their order of introduction to [B/D].)
• BUILD:  the data selection for the newly created element is defined to be the subset of observations for which there is a genuine observation on every data carrier in the definition part of the command. (DW this is about to change!)
• EXPORT:  all possible data
• LOOK:  all possible data for the data-carriers intended.
• PLOT:  and LOPLOT:  the first data carrier or element possessing data encountered. (If base notation is used for elements possessing data, elements are taken in their order of introduction to [B/D].)
• SCAN:  the first element encountered. (If base notation is used for elements possessing data, elements are taken in their order of introduction to [B/D].)
• SUMMARY:  all possible data for the data-carriers intended.

 

Missing value

 

Usage

• BD>control  missing =I where I is an integer.

• the default value is .

The missing  control is used to set an integer value to represent missing data cases. For example, to use to represent missing observations, issue the command

BD>control  missing  = -333

The control is used in two different ways. Firstly, when exporting data via the EXPORT:  command, all completely undefined cases are output as the missing value. Secondly, when data is being entered via a DATA:  command into [B/D], this value is used to mark missing values, and the program treats them thereafter as being undefined.