=== Definition ===

=== Practical Computations ===

=== How to run ===

This is module is include in MW version XXXX

In order to use this module You need to modify three files
   1. dbook.inc: changes the definitions accordingly of what you want to do.[[br]][attachment:dbook.inc dbook.inc]: Example of dbook.inc in order to plot three different variables.
   1. topgraph.f: defines the value to plot[[br]][attachment:topgraph.f topgraph.f]: Example of topgraph.f in order to have Mtt, Pt of the top quark and angular distribution of the top quark.
   1. Madweight_card.dat:

You should add the following option in the MadWeight _card
|| BLOCK  || variable name ||  example of value || explanation ||
|| mw_run  || histo ||  T ||  # creates histogram for value defined in topgraph.f ||
|| mw_run  || histo_by_events ||  T ||  # creates histogram for each events (Optional) False by default  ||

=== Kinematical fit ===

==== What is done ====
   * matrix element to one
   * mass fix at pole mass
   * integration on the transfer function

==== How to run ====
   * use the KIN_TEMP in ingrid: /home/fynu/omatt/scratch/MG_ME_DMTT/
   * copy that Template
   * ./bin/PassToMadWeight
   * configure the dbook.inc/topgraph.f (see above)
   * configure all the card
   * launch

=== Estimating error and deviation between theory experimental result ===

==== The principle ====

Usual estimation of the error on each bin based on the square root of the number of entry is not valid in the current scope due to the strong correlation between the bin entry. We then need a more evolve way to estimate the standard deviation.

Additionally, we want to be able to claim the validity of an experimental curves compare to it's theoretical couterpart. Due to the correlation, the usual chi-square method is not valid and we need to use an extended version based on the correlation matrix.

In consequence, in this section we will present two tools. The first one computes the standard deviation and the correlation matrix. it's base on the computation of the expected value/standard deviation/correlation of the value of one bin for an experiment containing a random number of experimental event following of poisson of parameter one. (this method was validate using some simple case)

==== Compute the error/sigma/correlation ====

requirement: python (at least 2.5) with numpy installed

program: [attachment:correlation.py.txt correlation.py.txt]: correlation.py.txt (note that you have to remove the .txt)

after the ''''''main'''''' == '''''name''''':

edit the init_dir (position of data)

edit the output_dir (where to write the output) then write the following line:

MW_correlation_mult(XXX, YYY, bin(VVV, WWW)) where
   1. XXX is the number of events in your directory
   1. YYY is the variable number
   1. VVV is the value of the first bin
   1. WWW is the width of a bin (we suppose that we have 50 bins) then python correlation.py

==== Estimate the chi-square ====

requirement: python (at least 2.5) with numpy installed

program: [attachment:dist.py.txt dist.py.txt]: dist.py.txt (note that you have to remove the .txt)

all modification to perform, set the end of the file

{{{
#!html

<div id="''mcePaste">if ''''''main''''''=='''''name''''':</div> <div id="''mcePaste">   obj2=Chi_carre('MW_mean','MW_error','MW_correlation',Nb_event)</div> <div id="''mcePaste">   chi''carre=obj2.compute_from_file('MW_SM_event',mode=5)</div> <div id="''mcePaste">   print chi''carre</verbatim></div>
}}}
'MW_mean','MW_error','MW_correlation' are the value computed for one event.

Nb_event is the number of event if your analyzed histogram.

'MW_SM_event' is the histogram files.

-- Main.OlivierMattelaer - 13 Oct 2009