[daip] AIPS handling of weights and flags

[Eric Greisen]

Eskil Varenius wrote:
> Dear AIPS support,
> I wonder about how AIPS handles weights and flags. Please direct me to 
> the appropriate place if this address is not the correct one. I've 
> searched the net, cookbook and explain-pages without a clearcut answer. 
> I would be very happy if someone could increase my knowledge around the 
> following questions.
> 
> What is the meaning of negative weights for visibilities? Is there a 
> general way in which such visibilities are handled throught AIPS?
> Suppose that I modify weights on some visibilities using WTMOD (or 
> ParselTongue) to -1. If I then use the task FIXWT to modify weights to 
> reflect the amplitude scatter of data, will this also modify the 
> negative weights? Or, will it only modify positive weights?

Negative or 0.0 weights are viewed as flagging the data and the 
visibility values for such data are seldom changed and never used.  I 
cannot promise what individual tasks might do, but the <= 0 status of a 
weight should never change.

> Finally I wonder about SPLIT. SPLIT can apply an FG-table. If I make a 
> single-source file using SPLIT from a multi-sourcefile, and chose to 
> flag data using the multi-sourcefile-FG-table, will SPLIT remove every 
> single flagged visibility? Or - since I don't see any FG table in the 
> produced single-source file - are some visibilites still transferred for 
> some reason but with modified weights? What about visibilites with 
> negative weigths, will these count as flagged when using SPLIT?
> 
When an FG table is applied, the calibration routines change the weights 
to - abs(weight) and if all weights in the visibility record (one 
baseline, one time but all IFs and all polarizations and all spectral 
channels) are <= 0, the record is discarded and will not appear in the 
output data set.  Thus some vis may appear with <= 0 weights (say you 
flagged LL but not RR for some time) in the output whilst others will 
disappear (you flagged both RR and LL at that time).  Visibilities with 
negative weights in the input set to SPLIT continue to have negative 
weights and will disappear (or not) depepnding on the other parts of the 
vis record.

Eric Greisen