Syntax
BD>var : v(S
where
This usage of the VAR: command is used to construct variance
matrices from linear combinations of other variance and/or correlation
matrices. The naming part indicates that the beliefs being specified are
the covariances for belief store number
The matrices in the linear combination are either variance matrices or
correlation matrices. For example, the syntax v(S
The matrices involved in the command must be conformable. You should be
careful when you use commands of this kind when the collections in the
definition part are neither the same as, nor disjoint to, the
collections in the naming part. This is because [B/D] assumes all
variance-covariance matrices to be symmetric and thus stores only part
of the matrix. Consequently, asymmetries in what you attempt may result
in unintended misspecifications. See §6.8.5 for an
illustration of this kind of problem. These problems do not arise when (1)
If an equation such as
,B ,B 
)=[(E )][v(S ,B 
,B 
)]
[(E )][r(S ,B 
,B 
)] 
(E 
)
... 
are the names of bases which must
already exist, 
are valid equations, and
are belief store numbers.
between the collection of
elements 
and the collection of elements 
. (If
then 
may be omitted.) The optional
coefficients in the linear combination are the equations 
which must be enclosed within round brackets.
,B ,B )
refers to the variance-covariance matrix specified between the
collection of elements 
and the collection of elements
for belief store number 
. (If 
then
may be omitted.) r(S ,B ,B ) refers to the
correlation matrix specified between the collection of elements
and the collection of elements 
for belief store
number 
. (If 
then 
may be omitted.)
Where a correlation matrix is indicated, the corresponding covariance
matrix is accessed and the correlations calculated temporarily.
is disjoint to 
and (2) 
, and similarly for any other pair
of collections appearing in the definition part. It is perfectly
possible for 
and 
as the definition
part is evaluated entirely before being stored.
is found without being the coefficient
of a succeeding matrix, a matrix of conformable dimensions of scalars
all equal to the value of the equation is added or subtracted
as necessary.