next up previous contents index
Next: Functional belief specifications Up: Declaring variances and covariances Previous: Direct numerical specification of

Problems of asymmetry

 

The beliefs that you specify over pairs and so forth must be symmetric: that is, for all X,Y we must specify tex2html_wrap_inline35332 . An illustration of the kind of problem that might arise is given below.

BD>element : a,b,c,d,e,f tex2html_wrap_inline33712

BD>base : x=a,c,d,e tex2html_wrap_inline33712

BD>base : y=b,c,d,f tex2html_wrap_inline33712

BD>var : v(2,x,y) tex2html_wrap_inline33712

BD*-2 -1 -1 -6 tex2html_wrap_inline33712

BD*-1 0 6 -5 tex2html_wrap_inline33712

BD*-1 5 0 -5 tex2html_wrap_inline33712

BD*-6 -5 -5 -6 tex2html_wrap_inline33712

This sequence of commands specifies a tex2html_wrap_inline35334 covariance matrix between the collections x and y. However, x and y are not disjoint - they both contain the elements c and d - and the actual specification mistakenly attempts to define tex2html_wrap_inline35336 and tex2html_wrap_inline35338 . An error is flagged whenever this kind of mistake occurs. This problem can arise also when using the VARDATA:  and MATMULT:  commands. 



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