This page is a tutorial explains everything related with systematic errors in highland:

Introduction

The flat tree is an intermediate level of data reduction, which can be used as input for any highland2 executable. It has the following properties

All variables stored in the flat tree are floats, integers and arrays of them
All variables are in the same directory
There is one entry per bunch unlike the oaAnalysis tree which has one entry per spill. Then bunching has to be done before filling the flat tree
Usually a preselection will be done. The application that creates the flat tree (see below) currently save only events with at least a FGD1 or FGD2 track and events wich are true numuCC, nueCC, antinumuCC or NC interactions in FGD1 and FGD2 volumes. This is just an example, since the preselection can be changed depending on the analysis.

All this makes the flat tree much smaller (4 GB in a single file against 110 GB in many files) and faster (>20 times) than using the oaAnalysis files.

The flat tree blocks are defined and filled in the Define and Write methods of the data classes (see DataClassesIO.cxx under highlandIO). The flat tree is ultimately created by the application RunCreateFlatTree under highlandIO, in which the blocks to be written and the preselection is done. This application is run as any other application in the framework:

RunCreateFlatTree.exe -o flattree.root oaAnalysis.list

As mentioned above, the application applies a pre-selection based on the true and reconstructed information in a spill/bunch. You can specify the requirements by editing the highlandIO.params.dat file in the $HIGHLANDIOROOT/parameters directory. You should ensure that the flat tree you create contains information for all the true signal events (so the analysis efficiency can be calculated correctly) as well as any events that pass loose reconstruction cuts (say, have a track in a TPC, if you're doing a Tracker-based analysis).

Once the flat tree is created, any analysis can be run using it as input:

RunNumuCCAnalysis.exe -o microtree.root flattree.root

The way the flat tree is constructed ensures that results should be identical to those obtained with oaAnalysis files.

Developers who extend the DataClasses: In order to keep the Flat Tree up to date you should:

propagate to the Define and Write methods (in DataClassesIO.cxx) any addition to the DataClasses members
convert back to the DataClasses any new variable in the Flat Tree. This should be done in the FlatTreeConverter under highlandIO

Parameters for FlatTree creation

Whether to enable the CT4/CT5 POT correction, and the DQ
corrections. These must be applied when creating the flat
tree, so that POT counting is correct. Applying the same
corrections in baseAnalysis when using FlatTree files as 
input will have no effect. IT IS STRONGLY RECOMMENDED THAT
YOU KEEP THESE CORRECTIONS ENABLED!
 < highlandIO.FlatTree.EnableIgnoreRightECalCorrection = 1 >
 < highlandIO.FlatTree.EnablePOTCorrection             = 1 >
 < highlandIO.FlatTree.EnableDQCorrection              = 1 >
 < highlandIO.FlatTree.EnableTPCdEdxDataCorrection     = 1 >
 < highlandIO.FlatTree.EnableTPCdEdxMCCorrection       = 1 >
 < highlandIO.FlatTree.EnableTPCExpecteddEdxCorrection = 1 >
 < highlandIO.FlatTree.EnableTPCPIDPullMCCorrection    = 1 >
 < highlandIO.FlatTree.EnablePileUpCorrection          = 1 >
 < highlandIO.FlatTree.EnableUseMuonAltMomCorrection   = 1 >
 < highlandIO.FlatTree.EnableMomResolMCCorrection      = 1 >
 < highlandIO.FlatTree.EnableFlipKinematicsCorrection  = 0 >
 < highlandIO.FlatTree.EnableToFCorrection             = 0 >
 < highlandIO.FlatTree.EnableMomRangeCorrection        = 0 >
 
Save only true tracks/vertices with recon association
 < highlandIO.FlatTree.SaveTruthWithReconAssociation = 0 > 
 
Save RooTracker information, which is needed for doing re-weighting
with T2KReWeight
 < highlandIO.FlatTree.SaveRooTracker = 1 > 
Save only the RooTracker vertices with link in the MiniTree  (ONLY FOR MiniTree)
 < highlandIO.FlatTree.FilterRooTracker = 1 > 
Enable/disable filling subdetector variables
 < highlandIO.FlatTree.SaveTPCInfo     = 1 >    
 < highlandIO.FlatTree.SaveFGDInfo     = 1 >    
 < highlandIO.FlatTree.SaveP0DInfo     = 1 >    
 < highlandIO.FlatTree.SaveECALInfo    = 1 >    
 < highlandIO.FlatTree.SaveSMRDInfo    = 1 >    
 < highlandIO.FlatTree.SaveTrackerInfo = 1 >    
Enable/disable the reading and flattening of different oaAnalysis folders
 < highlandIO.FlatTree.UseReconDirTrackerECal = 0 >
 < highlandIO.FlatTree.UseReconDirP0DECal     = 0 >
 < highlandIO.FlatTree.UseReconDirP0D         = 0 >
 < highlandIO.FlatTree.UseReconDirFGDOnly     = 0 >
Fill the P0DDataClasses
 < highlandIO.P0DDataClasses.UseReconDirP0DNew            = 0 >               Enable/disable reading the P0DRecon dir and filling the new P0DDataClasses
 < highlandIO.P0DDataClasses.UseReconDirP0DAlgoResult     = p0dRecon >        Take P0D vertices from this algo result
 < highlandIO.P0DDataClasses.AddGlobalTracksToP0DVertices = 1 >               Global tracks containing a P0D segments are added to the corresponnding P0D vertex
Preselection based on subdetector components 
 < highlandIO.FlatTree.UseTPC1    = 0 > 
 < highlandIO.FlatTree.UseTPC2    = 0 > 
 < highlandIO.FlatTree.UseTPC3    = 0 > 
 < highlandIO.FlatTree.UseFGD1    = 1 > 
 < highlandIO.FlatTree.UseFGD2    = 1 > 
 < highlandIO.FlatTree.UseP0D     = 0 > 
 < highlandIO.FlatTree.UseDsECal  = 0 > 
 < highlandIO.FlatTree.UseTrECal  = 0 > 
 < highlandIO.FlatTree.UseP0DECal = 0 > 
 < highlandIO.FlatTree.UseSMRD    = 0 > 
 < highlandIO.FlatTree.UseTPC1outOfBunch    = 0 >
 < highlandIO.FlatTree.UseTPC2outOfBunch    = 0 >
 < highlandIO.FlatTree.UseTPC3outOfBunch    = 0 >
 < highlandIO.FlatTree.UseFGD1outOfBunch    = 0 >
 < highlandIO.FlatTree.UseFGD2outOfBunch    = 0 >
 < highlandIO.FlatTree.UseP0DoutOfBunch     = 0 >
 < highlandIO.FlatTree.UseDsECaloutOfBunch  = 0 >
 < highlandIO.FlatTree.UseTrECaloutOfBunch  = 0 >
 < highlandIO.FlatTree.UseP0DECaloutOfBunch = 0 >
 < highlandIO.FlatTree.UseSMRDoutOfBunch    = 0 >
Preselection based on true interaction type
 < highlandIO.FlatTree.TrueVertex.NuNC        = 1 >
 < highlandIO.FlatTree.TrueVertex.AntiNuNC    = 1 >
 < highlandIO.FlatTree.TrueVertex.NumuCC      = 1 >
 < highlandIO.FlatTree.TrueVertex.AntiNumuCC  = 1 >
 < highlandIO.FlatTree.TrueVertex.NueCC       = 1 >
 < highlandIO.FlatTree.TrueVertex.AntiNueCC   = 1 >
Preselection based on true interaction detector
 < highlandIO.FlatTree.TrueVertex.TPC1    = 0 > 
 < highlandIO.FlatTree.TrueVertex.TPC2    = 0 > 
 < highlandIO.FlatTree.TrueVertex.TPC3    = 0 > 
 < highlandIO.FlatTree.TrueVertex.FGD1    = 1 > 
 < highlandIO.FlatTree.TrueVertex.FGD2    = 1 > 
 < highlandIO.FlatTree.TrueVertex.P0D     = 0 > 
 < highlandIO.FlatTree.TrueVertex.DsECal  = 0 > 
 < highlandIO.FlatTree.TrueVertex.TrECal  = 0 > 
 < highlandIO.FlatTree.TrueVertex.P0DECal = 0 > 
 < highlandIO.FlatTree.TrueVertex.SMRD    = 0 > 