[daip] Re: aips error analysis

[Eric Greisen]

Jon Mauerhan writes:
 > Hello,
 > I am interested in using the weights provided by the task PRTUV to compute 
 > errors in the amps & phases for each visibility in a uv dataset. There is very 
 > little information on how the weights provided by this task are generated. Can 
 > you offer some advice on how to translate these weights into actual amp & phase 
 > errors? Thanks for your help.

Weights are associated with each data point.  These are set by the
source of the UV data.  They can then be calibrated if that is
appropriate.  PRTUV prints the current values of these weights but
does not apply calibration.  UVPRT can calibrate weights (DOCAL = 2).

In principle, the weights are 1/sigma**2 in 1/Jy**2.  VLA data were
read into AIPS with the weights simply reflecting a count of integer
10's of seconds for the integration time.  Some users still do this,
although we encourage them to read weights that can be calibrated to
something like 1/Jy**2 (although it is not exact).  The integer
weights are proportional to integration time but not to uncertainties
since the different antennas, different IFs, etc will have different
noise levels.

Note that there are other sources of error than the receiver/front-end
temperature and so the noise estimates that include onnly that may be
seriously in error.

Eric Greisen