1 Using_Ntuples

  When talking about Ntuples in Mn_Fit most commands apply to Ntuples,
  scatter plots, series of points, tables and multi-dimensional
  histograms (including 2-D).  There are a number of commands
  available for manipulating them all.  See HELP NTUPLE for
  details. You can define cuts, which are either simple expressions
  (variable condition value), more complicated expressions (expression
  condition expression), or a COMIS function filename. These cuts can
  then be applied using the `NTUPLE PROJECT' or `NTUPLE PLOT' commands
  and you can project onto either a 1 or 2-dimensional histogram, a
  profile plot, a scatter plot or another Ntuple.  You can either
  specify what the binning should be for the resultant plot or let the
  program do it for you.  It is possible to use a variable as a weight
  or error on the weight for each point.  I have used this when
  measuring a 2-dimensional surface.  The measurements were stored in
  a table as `x,y,z,dx,dy,dz' and read in using `DAT_FETCH' with an
  `NTUPLE' card.

  It is also possible to plot Ntuples directly. This is most useful
  if you have read in a table of numbers (using the `DAT_FETCH' command
  for example) and want to plot one vs. the other without applying any
  cuts.

  If you specify a COMIS function as a cut (`CUT FILE' command), and the
  file does not exist a skeleton will be made. If you give the identifier
  of an existing Ntuple as part of the command then the variable names
  will be used in the skeleton, making it easy for you to specify your cuts.

  It is also possible to merge several Ntuples together, which have been
  created by separate jobs for example. You can call a COMIS subroutine
  for each event in the Ntuple (`NTUPLE SCAN') and can make whatever
  plots or calculations you wish there. You can select the events that
  you want to use for further analysis using the `NTUPLE FILTER' command.

  It should be trivial to adapt COMIS functions which have been used in
  PAW to Mn_Fit. Although KUIP vectors are not available, Mn_Fit
  registers and user variables are, so you can use these instead.

  Columnwise Ntuples (CWNs) are also supported. For details on how to
  book this sort of Ntuple see the HBOOK manual. The main advantage of
  them is the variable number of columns per event and that only the
  columns that are needed are fetched.  If you `SCAN' a CWN, all
  columns will be fetched.  `NTUPLE FILTER' does not yet work with
  CWNs. If you `PROJECT' then only those variables that are projected
  onto will be fetched. However I do not know which variables you want
  to access if you are using a COMIS function as a cut, so you have to
  give the command `SET NTUPLE VARIABLE' to list which variables are
  needed.

  Ntuple commands are still being developed and so there are likely to
  be additions and improvements to them in the future. Suggestions for new
  commands or how to make existing commands easier to use or more flexible
  are welcome.

