USAGE:
          urtapd @ <urtapd.ins>

 PURPOSE:
    Reads the dumps of urtapt (file unit 10: LSQ design matrix and data vector; file unit 12: Eigen-system)
    and runs subroutine SOLVI (SOLVI_P) on data arrays that are incremented with +1.0 in user-specified intervals.
    Thus, the dependence of the solution on specific data offsets can be studied.

    This may just be a beginning. Other types of signals may follow (the obstacle is that the design matrix
    is condensed, i.e. gaps of missing data have been closed; time series for superposition must be compatible
    with the irregularities that thus arise).
   
    urtapd is useful in the context of urtapm / urtap-merged problems, since the Feature table and the jd-series (Julian dates) are
    available.

INSTRUCTION FILE:

    Open file block
    Namelist &param
    Instructions
    /*
    
    Open file block:
10 ^ gmx-file          output from urtapt on unit 10
12 ^ evs-file          output from urtapt on unit 12
21 ^ weights-file      output from urtapm t/*.dw.ts
31 ^ dep.mc-file       computed: deviations of predictions     
32 ^ env.mc-file       computed: their envelope                
   q
    Namelist &param
-----------------------------------------------------------------------------------
Variable    type       meaning [default]
-----------------------------------------------------------------------------------
n_select    integer    How many parameters to vary [all]
i_select()  integer    which parameter [any single one]
                       (check with column number in urtap o/*.prl)
qdegf       logical    normalize output (unit 31 and 32) by degrees-of-freedom
weedlvl     real*8     flag as missing the data (before unweighting)
                       that exceeds weedlvl times RMS
idate(3)    integer    If no file is opened on unit 21,
                       specify the date  [2009,6,30] and ...
dt          real*8     ... sampling interval (in hours! *)) [5 s]    *) needs change to seconds
opt         char*16    experimental. Don't worry, the default '/e' is o.k.
nu          integer    number of eivenvalues to use [all]
sig         real*8     a forced scaling of the rms of the predicted series [1.0d0]
------------------------------------------------------------------------------------

    Instructions (no intermittent blank lines, neither before nor between nor after)
FF  i n                index range for adding a unit step to the input data: from i for n samples
FT  i j                index range for adding a unit step to the input data: from i to j
/*                     end-of-data symbol