scale
Change the scale factor that mutiplies the raw visibility weights.
EXAMPLE
-------

0>wtscale
Current weight scale factor is: 1
0>
0>wtscale 15
Changed weight scale factor to: 15
0>
0>wtscale wtscale()*2.3
Current weight scale factor is: 15
Changed weight scale factor to: 34.5
0>

PARAMETERS
----------

scale  -  The new scale factor to mulitply the raw visibility weights by.
          Note that increasing the weights corresponds to decreasing
          the apparent visibility amplitude uncertainties, and vice
          versa. In particular, to scale the error bars by a factor C,
          the argument of the wtscale command would have to be 1/C^2.

CONTEXT
-------
In UV FITS files each visibility is given a weight that is proportional
to the reciprocal of the associated visibility amplitude variance.
Unfortunately the constant of proportionality that is optionally given
in the same files, is rarely anywhere near the right magnitude, let
alone correct. The two adverse results are that 'vplot' displays
amplitude uncertainty error bars that are totally non-sensical, and
that the self-calibration commands report scaled Chi-squared values
that make it difficult to judge the significance of the fit.

The 'wtscale' command has been provided to enable one to change the
scale factor. In some cases it is possible to deduce the appropriate
weight scale factor by knowing where the data originated from, but in
other cases getting the "right" factor is unfortunately a matter of
guess work. Trial and error based on inspection of the error bars in
'vplot' is the only way to make this guess at the moment. Note 
that scaling the weights by X scales the error bars by 1/sqrt(X).

If the data-set is written back to a UV FITS file the weights will be
left un-scaled, but the accumulated scale factor will be recorded via
the AIPS mechanism of writing a HISTORY line of the form:

HISTORY AIPS WTSCAL=....

Both AIPS and Difmap recognise such lines, so if the file is later
read back into difmap or AIPS the weight scale factor will be corrctly
recovered. Note that since history lines are written incrementally,
there may be many history lines of the above form. The last such line
is the one actually used.

VLA DATA 
--------

The VLA writes weights equal to the number of 10s intervals in an
integration. The correct weight scale factor can be determined from
the equation for the expected noise variance:

 1/variance = (1e-26)^2 * Bw.t.(A.na.nc)^2
                          ----------------
                          2.0 * (k.Tsys)^2

Boltzmann's contant  :    k = 1.38066e-23 J/K
System temperature   : Tsys = 40K (typical for the VLA).
Observing bandwidth  :   Bw
Integration time     :    t = const * 10s
Antenna area         :    A = 490.8739 m^2
Antenna efficiency   :   na = 0.63
Correlator efficiency:   nc

The correlator efficiency depends on the VLA observing mode, but is
documented in the synthesis imaging workshop book.

Thus given that the raw data weights equal the number of 10 second
intervals in an integration, the correct value of wtscale is:

 wtscale = Bw * nc^2 * 1.5678e-4

Note that when used to combine two stokes parameters into a derived
polarization (eg. RR+LL->I), AIPS SPLIT doesn't combine the
weights correctly. If you are using data that have been processed in
this fashion, the above wtscale factor should be scaled by a factor of
2. If you combine RR+LL->I with 'select' in difmap there is no need
for this.