Scanning the effect of withdrawing information

For the third form of the syntax we assume that an adjustment has already taken place. We will assume that the collection to be adjusted is the base B, and that the currently fitted collections can be arranged as the (non-empty) collection D fitted explicitly as . Hence, we assume the scenario that would result if we issued the command

BD>adjust : [ B / D1 + D2 + ...+ Dn ]

The base scanned, F, must be one of the bases so added. That is, F must be D1 or D2 or ...or Dn.

The action of the SCAN:  command is now to evaluate the partial loss in resolution that would result from withdrawing F from the adjustment.

For example, suppose that the base being withdrawn, F, is . Then we assess the differences between the adjustments and .

The following information is available after such a command has been issued.

• the system adjustment uncertainty (essentially the standardised uncertainty remaining in the collection B having adjusted by D and then removing F), is stored in the operand scurvar . In the notation of [31, page 13,], the quantity stored is

  

• For each pair of elements in the base B, for example the pair , we store their adjusted covariance,

  

  This then can be accessed via the [B/D] operator scac  as scac  analagously to the [B/D] operator ac . For adjusted variances, we could also use the scav  operator in that scav  is equivalent to, and shorthand for, scac 

• By setting the scac  control to some valid belief store, the potential adjusted covariances

  

  may also be directed to a belief store. Every time the SCAN:  command is issued, the adjusted covariances are then output to this belief store. This facility works independently of the availability of the adjusted covariances via the scac  operator.

• For each element in the base B, we store the adjusted expectation,

  

  This then can be accessed via the [B/D] operator scae  as scae  analagously to the [B/D] operator aex .

• By setting the scae  control to some valid expectation store, the potential adjusted expectations

  

  may also be directed to an expectation store. Every time the SCAN:  command is issued, the adjusted expectations are then output to this store. This facility works independently of the availability of the adjusted expectations via the scae  operator.

As an example, consider the fragment of code given in Figure 9.5. Here, we scan for the effect of removing F from the adjustment of B by D. Assume that the base B contains elements , , and . The SCAN:  command itself performs the partial withdrawal of information, and then there are various lines of output which show how we access the results of the command.

 
Figure 9.5:  Scanning withdrawals of information

Similar information to this (albeit in terms of resolutions) can be obtained by using the arcout  operator following an adjustment.