.. _MADFlagger.json#/:

MADFlagger
==========

Automatic flagging in time/freq windows based on Median Average Deviation (MAD) filtering. Filter to filter on baseline and/or channel (only the given baselines/channels are kept).  The reader step has an implicit filter.


.. _MADFlagger.json#/inputs:

inputs
******


.. _MADFlagger.json#/inputs/type:

type
++++

:type: ``string``

:default: ``madflagger``

:doc: Case-insensitive step type; must be 'madflagger' (or 'madflag') `.`


.. _MADFlagger.json#/inputs/applyautocorr:

applyautocorr
+++++++++++++

:type: ``boolean``

:default: ``False``

:doc: True means that the MADFlagger is used on the auto-correlations only.  The resulting flags are applied to the cross-correlations, thus data  are flagged where the corresponding auto-correlations are flagged. An error is given if set to True, while the MS does not contain  auto-correlations `.`


.. _MADFlagger.json#/inputs/blmax:

blmax
+++++

:type: ``integer``

:doc: Maximum baseline length (in meters). It is similar to minimum `.`

:default: ``1e30``


.. _MADFlagger.json#/inputs/blmin:

blmin
+++++

:type: ``integer``

:doc: Minimum baseline length (in meters). Only baselines with a length >= this minimum are flagged.  If ``applyautocorr=true``, the autocorrelations are applied to the  matching baselines only.

:default: ``-1``


.. _MADFlagger.json#/inputs/correlations:

correlations
++++++++++++

:type: ``integer?``

:default: ``[]``

:doc: The correlations to use in the flagger; an empty vector means all.  They are handled in the order given; if the flagging criterium  holds for one correlation, the other correlations are not tested  anymore. So if one knows that most RFI is found in YY, then in  XX and finally some in XY and YX, the vector should be ``[3,0,1,2]``  because it makes the program run faster. Note that the statistics  printed at the end show how many flagged data points have been  found per correlation.


.. _MADFlagger.json#/inputs/count.path:

count.path
++++++++++++++

:type: ``string``

:default: ``""``

:doc: The directory where to create the flag  percentages table. If empty, the path of the input MS is used `.`


.. _MADFlagger.json#/inputs/count.save:

count.save
++++++++++++++

:type: ``boolean``

:default: ``False``

:doc: If true, the flag percentages per frequency are saved to a table with  extension ``.flagfreq`` and percentages per station to a table with extension ``.flagstat``.  The basename of the table is the MS name (without extension) followed by the stepname and extension `.`


.. _MADFlagger.json#/inputs/freqwindow:

freqwindow
++++++++++

:type: ``integer``

:default: ``1``

:doc: Number of channels in the median box. If not odd, 1 is subtracted.  It is silently reduced if exceeding the actual number of channels. In a way similar to ``threshold`` it can be made baseline length  dependent.


.. _MADFlagger.json#/inputs/step_name:

step_name
+++++++++

:type: ``string``

:default: ``madflagger``

:doc: unique name for the step `.`


.. _MADFlagger.json#/inputs/threshold:

threshold
+++++++++

:type: ``float``

:default: ``1``

:doc: The flagging threshold that can be baseline dependent. It can be any (`TaQL-like <http://www.astron.nl/casacore/trunk/casacore/doc/notes/199.html>`__) expression that evaluates to a float.  In the expression the variable 'bl' can be used which is the  baseline length (in meters). In this way the value can be made  baseline dependent. For example: ``iif(bl<100, 0.5, iif(bl<500, 0.75, iif(bl<1000, 0.9, 1)))`` defines the threshold between the baseline lengths 100, 500,  and 1000 meter.


.. _MADFlagger.json#/inputs/timewindow:

timewindow
++++++++++

:type: ``integer``

:default: ``1``

:doc: Number of times in the median box. If not odd, 1 is subtracted.  It is silently reduced if exceeding the actual number of time slots. In a way similar to ``threshold`` it can be made baseline length  dependent.