next up previous contents index
Next: Deleting functionally specified variances Up: Functional belief specifications Previous: An introduction to components

Declaring functional variances and covariances


tex2html_wrap_inline33790 tex2html_wrap_inline33790 Syntax

BD>fvar : v(S, X [IL tex2html_wrap_inline33794 ], Y [IL tex2html_wrap_inline33798 ]) = E tex2html_wrap_inline33712

where S is a belief store number, X and Y are the names of components or elements, tex2html_wrap_inline35376 are index lists, and E is an equation, possibly involving indices from either index list.

tex2html_wrap_inline33806 tex2html_wrap_inline33806

The FVAR:  command is used to specify variances and covariances functionally for components and between components and elements. At least one of X and Y should be a component, but one may be an element. The functional belief being defined is expressly limited to belief store number S, and will overwrite any functional belief expressed for the same pair of quantities for the same belief store. However, further functional beliefs could be expressed about the same pair of quantities for different belief store numbers.

The index list comprises single alphabetic characters, preceded by the `.' symbol. An alphabetic character must not appear more than once in the entire naming part, and the number of indices per component is currently limited to 8.

The equation E will usually relate to the as yet unknown indices of the naming part. As an example, consider the quantities tex2html_wrap_inline35378 discussed in (6.1) above, and suppose that we wish to specify the variance-covariance structure for belief store number 1 for these quantities. We could use the following command to achieve this:

BD>fvar : v(1,eta.i,eta.j)=(.i=.j)+(.i<>.j)/(.i+.j) tex2html_wrap_inline33712

 



David Wooff
Wed Oct 21 15:14:31 BST 1998