The first card in the a priori input stream to PEP is a title
card.  Then come one or more a priori input groups.  An input
group begins with a namelist (NMLST4) which has as members:
N - I*4 - the number of parameters for which there are a priori
values or sigmas.
DIAGON - L*4 - .true. if the input B matrix is in diagonal form,
otherwise matrix is assumed to be lower diagonal half.
COVAR - L*4 - .true. if the input B matrix is a covariance matrix,
otherwise assumed that matrix is coefficient (or
information) matrix.
APEST - L*4 - .true. if an a priori estimate vector is included in
this group, otherwise assumed none such.
SQRTB - L*4 - .true. (.and.diagon) B is the sqrt of the required
value, i.e., either sigma or 1/sigma
OFFSET - L*4 - .true. if a priori estimates are of offset from
first parameter in packet.  In this case, the first
parameter name is a reference parameter, others are
offsets.  The X matrix may contain actual values, or X(1)
may be zero with offsets in X(2-N); only the differences
are significant.  The B matrix may contain sigma's (if
COVAR is .true.) or 1/sigma's (if COVAR is false); SQRTB
is used, APEST and DIAGON are ignored.
Note: offset a priori information generates a singular
a priori matrix.  For this reason, a priori nominals are
not computed, and iteration is prohibited.
W - R*8 - weight factor by which the input matrix (coeficient or
covariance) is multiplied.  If zero, the matrix is not
used -- NAMRED and MATCH are not called.
B - R*8 - input B matrix whose form, content, and dimension are
described above.
X - R*8 - input a priori parameter values (if APEST .true.) for
parameters whose sigmas are in the B matrix.

The parameters for which a covariance or coefficient matrix and a
priori estimates are input are indicated by name cards immediately
following the namelist.  One parameter name goes on each card and
the names are in the same order as the values in the X and B
matrices.  Most parameters have two part names; these are described
in subroutine NAMPRM.  The names can be placed on the name cards in
free field format, at least one blank separating the first part
from the second.  The name cards have to be ended by a blank card.
Any number of groups of namelists and names can be input and the
program stops looking when a NMLST4 with N = 0 is found.