The following results are available under the following scenario:
suppose that base has been adjusted initially by base
and then partially by
. We will
use the notation
to mean the collection with
excluded.
Suppose
that we might observe
and
(otherwise,
data-related results are not available). For the following results we
need to have issued the ADJUST: command with either the
noinfluence control switched off, or with the influence
option switched on. For example, we might have issued the commands
where and
are the names of bases. Typically, the
results we are about to describe are used to label the arcs of influence
diagrams. These labels can require some time and effort to calculate,
and so we don't normally bother - we can switch off the noinfluence
control to force their calculation if we need them. Under these
conditions, the following results are available.
Usage
Under the conditions described above,
the partial resolution in the base B resulting from adding
Under the conditions described above, and when data is available,
This is the observed size of the partial adjustment, divided by its
expected size, being the trace of the corresponding partial resolution
matrix, resulting from adding
Under the conditions described above,
the partial loss in resolution due to withdrawing
Under the conditions described above, and when data is available,
the observed size of the partial loss, divided by its expected
size, being the trace of the corresponding partial resolution matrix,
due to withdrawing
Under the conditions described above, and when data is available,
returns
as well as D to the adjustment. This value remains until overwritten,
so using this quantity if the above conditions are not satisfied can
give unpredictable results. The arcin operator thus is a measures
of the worth of the base
as a solitary carrier of
information. The corresponding operator arcind
gives the corresponding size ratio as a diagnostic for this quantity
when data is available.
Usage
returns
as well as D to the
adjustment of B. This value remains until overwritten, so using this
quantity if the above conditions are not satisfied can give
unpredictable results. The corresponding operator arcin gives
the partial resolution used in the divisor.
Usage
returns
from the overall
adjustment. This value remains until overwritten,
so using this quantity if the above conditions are not satisfied can
give unpredictable results. The arcout operator thus is a measures
of the worth of the base
in supplying information about B
that no other source gives. The corresponding operator arcoutd
gives a diagnostic for this quantity when data is available.
Usage
returns
from the adjustment.
This value remains until overwritten,
so using this quantity if the above conditions are not satisfied can give
unpredictable results.
The corresponding operator arcout gives the partial loss in resolution
used in the divisor.
Usage
returns
, the path
correlation between the bearing for the node
and the
aggregated bearing for the remaining nodes being used for the
adjustment. Zero is returned if the path correlation is undefined. See
[31, section 5.5,]. This measures the observed degree of
concordance between the information sources
and
. The
value remains until overwritten, so using this quantity if the above
conditions are not satisfied can give unpredictable results.