[daip] weight averaging in UVPLT

[Dan Homan]

Hi,

I was plotting data weights with UVPLT, and I think I found a problem.  I 
was using data from the VLBA with full stokes (RR, LL, RL, LR), and I was 
interested in the weights of a particularly high point.

In LL this point had a weight of    3.6 kilo  (1/Jy^2)

In RR this point had a weight of   80   kilo  (1/Jy^2)

In "I" this point had a weight of  42   kilo  (1/Jy^2)

So the "I" weight was clearly just the average of the RR and LL weights, 
but I don't think this is quite right.

Because weight = 1/error^2, I think the correct error propagation would be

error_I  =  0.5*sqrt(error_RR^2 + error_LL^2)

so  weight_I =  4*(weight_RR*weight_LL)/(weight_RR + weight_LL)    

which would give an "I" weight in the above example of 14 kilo (1/Jy^2) 

I suppose it is possible that this problem also affects other tasks which 
combine antenna correlations to form Stokes parameters (IMAGR, CALIB for 
aparm(3) = 1, etc.), but I couldn't think of a simple way to test for it.

 - Dan

-- 
Daniel C. Homan, Jansky Fellow                            
National Radio Astronomy Observatory, Charlottesville, VA
e-mail: dhoman at nrao.edu, phone: (434) 296-0286