d. white  april 1974  subroutine filtin
r. goldstein march 1975  modified to allow prdict to use filtered
solutions
paul macneil january, 1978 modified for iteration
modified for *command july 1978  r.b. goldstein
modified for split into filtin, filtpn, filtep
z. goldberg   march 1980

delta - r*8 - length of time in one epoch, expressed as
jd.fract
sweigh - r*8 - weight factor for individual s-matrices
(default is 1)

lfilt(nepoch) - i*2 - if lfilt(i).ne.0, do solution
for epoch i, if any lfilt(i).ne.0  where i<nepoch
then by implication a backwards filtering operation
takes place.
lfilt(i)>0  for which i is smallest defines the pep
internal saved solutions  for iteration purposes

lprdct(nepoch) -i*2- if lprdct(i).eq.0  do not do a prdict
for that epoch. if lprdct(i).eq.1 do a time span prdict
for that epoch. the time span is centered at the epoch
center and of a width equal to delprd

fict(50) - i*2 - filter controls
fict(1) = 0  nothing
fict(1) = 1  print propagated info matrix each epoch
fict(1) = 2  print propagated rhs  matrix each epoch
fict(1) = 3  print both

fict(3) =-1  no cleanup nor error matrix print
must use fict(3) = -1 because of changes for extended
precision     pem  june, 1979
fict(3) = 0  only error matrix print
fict(3) = n  n iterations of inverse cleanup in addsmr
with error matrix print

fict(4) = 0  no scaling for dpinv nor printout
must use fict(4) = 0 because of changes for extended
precision     pem  june, 1979
fict(4) = 1  print unscaled h matrix and inverse
fict(4) = 2  scaling by 1/sqrt(max(row))*h*
1/sqrt(max(col))
fict(4) = 3  print scaled and unscaled h and h inverse

fict(6)=0  do not supress printout in analyze link when
using filter
fict(6)=1  supress all but the last page of printout in
analyze when using filter..also supresses
jout data set

fict(7)=0     do not save normal equations for any epoch
fict(7)>0     save normal equations for this epoch
fict(7)<0     save normal equations counting backwards
from last epoch. the usual situation would be
fict(7)=-1 , filtering forward only, and
saving normal equations from last epoch.

insne - i*4 - input saved normal eqn data set (created in
nrmict stage)
filter - i*4 - filter data set (created in form stage)
outsne - i*4 - output saved normal eqn data set (created
in use stage)
smat - i*4 - smearing matrix data set (input to pep)
iconof - i*4 - initial condition (really, process
noise parameter) offset data set
wzero - i*4 - filter epochs & initial conditions (pnp's)
before first integration