The likelihood ratio for candidate j for a given source i is (de Ruiter et al. 1977 [2],Prestage & Peacock 1983 [4], Wolstencroft et al. 1986 [5],Irwin 1996 [7]):
with the general expressions
and
such that, as in [4]
where is defined as
In equation ,
may be
which depends on the
magnitude of the candidate.
Note that a number of authors (Benn 1983 [3], [5],
Sutherland & Saunders 1992 [6]) have included the
factor Q (a priori probability that an
identification exists above the plate limit) in the expressions for
and hence
. How ever this factor Q is unknown and has
to be guessed, which goes against the idea that a likelyhood ratio should be
derived from the data and make no a priori assumptions.
For this reason we chose no to include it, although it (or rather
, which is very similar to it) appears in the definition of
reliability (see