Review of [B/D] usage

To this point we have shown you how to start and stop [B/D]. We have used the following commands for basic organisation and input of a problem:

ELEMENT: 

to introduce uncertain quantities and their expectations;

BASE: 

to organise quantities into bases;

VAR: 

to specify beliefs in the form of variances and covariances for the uncertain quantities;

DATA: 

to specify data for the observable quantities.

To check upon our inputs using these commands, we used the LOOK:  command with several different arguments:

e 

for elements;

b 

for bases;

v1 

for variances;

r1 

for correlations;

d 

for data.

When our specifications were completed, we issued the ADJUST:  command, [B/D]'s most important command, which performs the general adjustment of one belief structure by another. The SHOW:  command was then used to extract various information about the adjustment. In particular, the following display options were used:

• e  and e*  to see the adjusted expectations and standardised adjusted expectations;
• v  and v+  to see adjusted variances for our quantities;
• a  and a+  to see the evaluations of the adjusted expectations for the particular data that we observed;
• cd+  to see the canonical directions and their resolutions;
• cda+  to see evaluations of the adjusted expectations for the canonical directions;
• b  and b*  to see the bearing and standardised bearing, summarising the observed adjustment.

In addition, we introduced the notion of an assignment meaning a qauntity in [B/D] being represented as a linear combination. We saw how the canonical directions and the bearing can be retained as assignments by using the KEEP:  command with arguments cd  and b  respectively; and we saw how they can be examined by using the LOOK:  command with argument a .

We showed you how to use the BUILD:  command to transform assignments into elements, and we constructed the canonical directions and the bearing from their representations as assignments into elements which became usable for further [B/D] analysis.

Finally, we considered some ways in which the [B/D] language is used to determine partial adjustments of belief, and we used the following options to see results specific to partial adjustments:

pv  to see a summary of information about the effects partial adjustment on uncertainties.

pcd+  to see the partial canonical directions and their resolutions.

pc  to display the path correlation.