The peculiar features in the spectral flux distribution of CP2 stars
cause deviations from the values of photometric indices of normal stars.
As previous authors have already
mentioned, the blanketing is increased in CP2 stars, so the *m*_{1}
index is larger than for normal stars
(Strömgren 1963, 1967; Cameron 1967)
and the hottest stars do not satisfy the [*u*-*b*] vs. *m*_{1} relation
although they satisfy the [*u*-*b*] vs. relation
(Strömgren 1966). On the other hand, the Balmer jump
decreases, so *c*_{1} decreases as well
(Preston 1975; Gerbaldi et al.
1974). The hottest CP2 stars look bluer
than normal stars and the coolest are too red
(Wildey et al. 1962). The index also decreases
(Hauck 1975).
Henry (1969) showed that the peculiar A0-A3 stars are below the relation
for normal stars in the Strömgren (*b*-*y*) vs. *a* diagram.

Most of the authors used *m*_{1} vs. *c*_{1} or *m*_{1} vs. (*b*-*y*)
diagrams (Cameron 1967;
Maitzen 1976; Hauck 1975;
Adelman et al. 1995,
among others) to separate peculiar from normal stars.
However, since all Strömgren-Crawford indices are altered by the
peculiarities, we make use of the whole photometric information concerning
peculiarity for the sake of the discrimination.

The method used was the *Multiple Discriminant Analysis* also called
the *Canonical Discriminant Analysis* (hereinafter MDA; see for example
Murtagh & Heck 1985). This method analyzes *g* populations, formed by
*n*_{g} individuals, described by *q* variables. The *g* populations are
represented along canonical orthogonal axes maximizing the spread of the
means of the populations and restraining their compactness.
The MDA is invariant under linear transformations of the variables
and it takes into account the correlation between them.
In our case, there are two populations (the normal stars and the peculiar stars)
and the observed variables are the Strömgren-Crawford photometric indices.
One axis is enough to represent two populations. We denote
*p* the coordinate associated with this axis.