Flagging Visibilities

In Miriad, flagging a correlation means you set a bit in a mask. There are no explicit weights like in AIPS or ascii flagging tables. When you list visibilities with uvlist (see below), flagged correlations are indicated by an asterisk to the right of the phase.

There are four main flagging tasks in Miriad:

• blflag - a plot-based interactive flagger. This normally plots one baseline at a time, with a variety of possible axes (e.g. time against amplitude), and you click on bad data to flag it out. This is similar to tha AIPS IBLED task.
• tvflag - a ``TV''-based, interactive flagger. This displays one baseline at a time on a ``TV'' display, with the axes of the display being channel number and time. You select regions of bad data in the display. This is similar to AIPS task SPFLG.
• pgflag - a ``PGPLOT''-based, interactive flagger. This displays one baseline at a time on a ``PGPLOT'' display, with the axes of the display being channel number and time. You select regions of bad data in the display. This is very similar to the tvflag task, except it can be run on any X display with PGPLOT installed, whereas tvflag can only be run on displays with 8-bit colour depth.
• uvflag - a non-interactive general flagger, where you specify the data to be flagged by the select keyword. This is like the AIPS task UVFLG.

There are a number of other flagging tasks which occupy some niche which are described in the next section.

Although there is some personal taste involved, a recommended scheme for flagging is as follows:

• For 20 and 13 cm observations it is strongly recommended that you start the flagging process with pgflag. Narrowband interference is common at these wavelengths, and pgflag is a good tool at finding and flagging this interference. pgflag is conveniently performed on a multi-source file, straight after the data are loaded into Miriad.
• Using blflag is generally quick compared to pgflag, and some forms of bad data are more apparent with blflag. However, bad data is not that apparent when attempting to use it with datasets containing multiple sources. We recommend that you split the data into single source datasets (using uvsplit) before using blflag. Using blflag is recommended even if you use have already used pgflag on your data, because blflag is relatively quick and painless.
• Normally blflag will apply any calibration tables to the data before generating its plot. Thus it is just as useful after calibration as before. It is useful to look at calibrators using blflag both before and after solving for calibration parameters. With the program source, you will probably only use blflag after you are happy with the calibration.