configMinimizerInterface

< back to parent (FitterEngine)

The minimizer interface can be configured to use either an MCMC to estimate the Bayesian posterior, or a maximum likelihood optimization to find frequentist estimators for the parameters (i.e. the maximum likelihood best fit point), and the covering region (i.e. the Covariance at the best fit point). The maximum likelihood optimization is controlled using the RootMinimizer object, while the MCMC is produced using the SimpleMcmc object. Both objects are configured using the configMinimizerInterface, and are described separately.

Config options for both minimizers

minimizerConfig Options	Type	Description	Default
type	string	Choose the SimpleMcmc or RootMinimizer	RootMinimizer
checkParameterValidity	bool	Turn on parameter validity checks (in physical, in domain, etc)	false
showParametersOnFitMonitor	bool	Display fit parameter parameter values on the monitor	false
enablePostFitErrorFit	bool	Apply error evaluation after the minimization (may not be implemented)	true
maxNbParametersPerLineOnMonitor	int	Number of parameters on a monitor line	15
useNormalizedFitSpace	bool	use fit parameter interface to provide prior mean at 0 and stddev at 1	true
writeLlhHistory	bool	write a ttree registering all the llh evaluations	false

The rest of the fields are specific to the specific minimizer type being used.

• The ROOT Minimizer provides a maximum likelihood optimization to find the best fit point of the likelihood and provides ways to calculate the Hessian at the best fit point.

• The Simple MCMC generates an MCMC chain of points drawn from the likelihood that characterizes the multi-dimension likelihood distribution (mostly used in Bayesian analysis).