Next: Coherence and data consistency
Up: Preparing adjustments
Previous: Organising data
We have pointed out elsewhere (see, for example,
§6.6.3) that there may be alternative
variance-covariance and expectation specifications for the same
collection of elements, and these alternatives will be located in
different belief and expectation stores. When we carry out an adjustment, we
anticipate by default that all relevant variance-covariance
specifications are in belief store number one, but this need not be
the case. Indeed, for exchangeable adjustments this is often not the
case.
As far as the expectation specifications are concerned, only the
current default expectation store is ever referenced; the remainder
are ignored. The default expectation store can be changed by using the
e argument to the CONTROL: command.
Suppose that we gather all of the information sources into the base D, and suppose that the sample size is unity.
The information required by the adjustment is as follows:
- The expectations and . These will have been specified
earlier, when the elements were defined using the ELEMENT:
command, and might since have been
modified by using the E: command.
- Possibly, observations on D. If these are not available, no
results for the observed adjustment are available.
- The prior variances specified over the collection being adjusted,
. By default these variances are assumed to be present in belief
store number one. If they are elsewhere, this can be changed by using the
priorvar control as the adjustment assumes that the prior variance
matrix over B can be found in the belief store pointed to by this
priorvar control.
- The prior variances specified over the collection used for the
adjustment, . By default these variances are assumed to be
present in belief store number one. If they are elsewhere, this can be
changed by using the infovar control as the adjustment assumes that
the prior variance matrix over D can be found in the belief store
pointed to by this infovar control.
- The prior covariances specified between the collection being
adjusted and the collection used for the
adjustment, . By default these covariances are assumed to be
present in belief store number one. If they are elsewhere, this can be
changed by using the betweenvar control as the adjustment assumes that
the prior covariance matrix between B and D can be found in the
belief store pointed to by this betweenvar control.
Next: Coherence and data consistency
Up: Preparing adjustments
Previous: Organising data
David Wooff
Wed Oct 21 15:14:31 BST 1998