UNBIAS

Program UNBIAS is used to correct for a positive bias that  exists  in
low  signal-to-noise-ratio  amplitude  data, since the noise is always
positive.  To run the program,  use  the  command  UNBIAS  defined  in
VLB:LOGIN.COM:

        $ UNBIAS

The program then asks four questions:  version of PHASOR, input  file,
output file, and integration time to be used.

     The integration time is the length of interval, in seconds, which
will  be  used to determine the average signal to noise ratio for each
interval within the data.  The integration time should be at least  as
long  as the averaging time (if any) you plan to use for the data, but
may be considerably longer if there is not a lot of structure  in  the
data.   The  longer  the  integration  time,  the  more  accurate  the
correction will  be,  up  to  the  limit  where  the  SNR  is  varying
appreciably due to structure in the source.

     Because the required correction is a function of SNR, the program
then  temporarily averages the amplitude data and the amplitude errors
to determine an average  SNR  for  each  time  interval.   An  empical
correction  (in the form of a polynomial, determined by Richard Simon)
is then applied to the UN-averaged data.   In  practice,  the  maximum
correction  is  as much as ~40% for extremely weak data.  If less than
about 60% of the data has converged in a given interval, the  data  is
too  weak  to calibrate, and you should perhaps toss it onto the waste
pile of scientific effort.  However, >~20% convergence does  mean  you
may have a weak detection.

COMMENTS ON THE CALIBRATION OF VLBI DATA

The following  is  a  logically  consistent  (and  hopefully  correct)
procedure for calibrating VLBI data, as of 4-July-1981:

     1.  Run VLBEDIT, to edit the correlation coefficients  to  remove
         the obviously "bad" amplitude data points.

     2.  If necessary, that is, if you were unwise and used PHASRB  to
         PHASOR  your  data  rather  than FPHASRB, correct for partial
         convergence in the data by using Roger Linfield's polynomial.
         This  is  only  necessary  for SNR's <~4, and MUST be done by
         hand.  See the memo written by RPL.

     3.  Run UNBIAS.

     4.  If you used FPHASRB, but wish the error bars to have the same
         magnitude  as  the  old  PHASRB  would  have  generated, then
         correct the error bars in the data set using VLBEDIT and  the
         following parameters:  AMP ERRMULT=0.86 INT=1,X where X is >=
         the no.  of records in the dataset.

UNBIAS                                                            Page 2


     5.  Run CAL (q.v.), to change data to flux density.

     6.  Verify calibration using UV crossing points, etc.

     7.  Run VLBEDIT, to correct for telescope mis-pointing, etc.

     8.  Run AVERAGE, if desired, to average data incoherently.

Note that several iterations of 5, 6, and 7 may be needed.  I urge you
to use command procedures to do all of the above!


                                               R.S. Simon, 1982 July 4