class RooStats::BayesianCalculator: public RooStats::IntervalCalculator, public TNamed

 default constructor

 Constructor from data set, model pdf, parameter of interests and prior pdf
 If nuisance parameters are given they will be integrated according either to the prior or
 their constraint term included in the model

 Constructor from a data set and a ModelConfig
 model pdf, poi and nuisances will be taken from the ModelConfig

 destructor

 clear all cached pdf objects

 set the model to use
 The model pdf, prior pdf, parameter of interest and nuisances
 will be taken according to the model

 Build and return the posterior function (not normalized) as a RooAbsReal
 the posterior is obtained from the product of the likelihood function and the
 prior pdf which is then intergated in the nuisance parameters (if existing).
 A prior function for the nuisance can be specified either in the prior pdf object
 or in the model itself. If no prior nuisance is specified, but prior parameters are then
 the integration is performed assuming a flat prior for the nuisance parameters.
 NOTE: the return object is managed by the class, users do not need to delete it

 Build and return the posterior pdf (i.e posterior function normalized to all range of poi)
 Note that an extra integration in the POI is required for the normalization
 NOTE: user must delete the returned object

 return a RooPlot with the posterior  and the credibility region
 NOTE: User takes ownership of the returned object

 set the integration type (possible type are) :
 1D integration ( used when only one nuisance and when the posterior is scanned):
    adaptive , gauss, nonadaptive
 multidim:
       ADAPTIVE,   adaptive numerical integration
                    The parameter numIters (settable with SetNumIters) is  the max number of function calls.
                    It can be reduced to make teh integration faster but it will be difficult to reach the required tolerance
       VEGAS   MC integration method based on importance sampling - numIters is number of function calls
                Extra Vegas parameter can be set using  IntegratorMultiDimOptions class
       MISER    MC integration method based on stratified sampling
                See also http://en.wikipedia.org/wiki/Monte_Carlo_integration for VEGAS and MISER description
       PLAIN    simple MC integration method, where the max  number of calls can be specified using SetNumIters(numIters)

 Extra integration types are:

    TOYMC:
       evaluate posterior by generating toy MC for the nuisance parameters. It is a MC
       integration, where the function is sampled according to the nuisance. It is convenient to use when all
       the nuisance are uncorrelated and it is efficient to generate them
       The toy are generated by default for each  poi values
       (this method has been proposed and provided by J.P Chou)
    1-TOYMC  : same method as before but in this case the toys are generated only one time and then used for
       each poi value. It can be convenient when the generation time is much larger than the evaluation time,
       otherwise it is recoomended to re-generate the toy for each poi scanned point of the posterior function


    ROOFIT:
       use roofit default integration methods which will produce a nested integral (not reccomended for more
       than 1 nuisance parameters)

 if type = 0 use default specified via class IntegratorMultiDimOptions::SetDefaultIntegrator

 Compute the interval. By Default a central interval is computed
 and the result is a SimpleInterval object.
 Using the method (to be called before SetInterval) SetLeftSideTailFraction the user can choose the type of interval.
 By default the returned interval is a central interval with the confidence level specified
 previously in the constructor ( LeftSideTailFraction = 0.5).
  For lower limit use SetLeftSideTailFraction = 1
  For upper limit use SetLeftSideTailFraction = 0
  for shortest intervals use SetLeftSideTailFraction = -1 or call the method SetShortestInterval()
 NOTE: The BayesianCalculator covers only the case with one
 single parameter of interest
 NOTE: User takes ownership of the returned object

 Returns the value of the parameter for the point in
 parameter-space that is the most likely.
  How do we do if there are points that are equi-probable?
 use approximate posterior
 t.b.d use real function to find the mode

 internal function compute the interval using RooFit

 internal function compute the interval using Cdf integration

 approximate posterior in nbins using a TF1
 scan the poi values and evaluate the posterior at each point
 and save the result in a cloned TF1
 For each point the posterior is evaluated by integrating the nuisance
 parameters

 compute the interval using the approximate posterior function

 compute the shortest interval

 constructor

 specify the parameters of interest in the interval

 specify the nuisance parameters (eg. the rest of the parameters)

 Set only the Prior Pdf

 set the conditional observables which will be used when creating the NLL
 so the pdf's will not be normalized on the conditional observables when computing the NLL

 set the size of the test (rate of Type I error) ( Eg. 0.05 for a 95% Confidence Interval)

 set the confidence level for the interval (eg. 0.95 for a 95% Confidence Interval)

 Get the size of the test (eg. rate of Type I error)

 Get the Confidence level for the test

 set the fraction of probability content on the left tail
 Central limits use 0.5 (default case)
 for upper limits it is 0 and 1 for lower limit
 For shortest intervals a negative value (i.e. -1) must be given

 set the Bayesian calculator to compute the shorest interval (default is central interval)
 to switch off SetLeftSideTailFraction to the rght value

 set the precision of the Root Finder

 use directly the approximate posterior function obtained by binning it in nbins
 by default the cdf is used by integrating the posterior
 if a value of nbin <= 0 the cdf function will be used

 set the number of iterations when running a MC integration algorithm
 If not set use default algorithmic values
 In case of ToyMC sampling of the nuisance the value is 100
 In case of using the GSL MCintegrations types the default value is
 defined in ROOT::Math::IntegratorMultiDimOptions::DefaultNCalls()

 force the nuisance pdf when using the toy mc sampling