DefaultLatentModel< copulaPolicy > Class Template Reference

Default event Latent Model. More...

#include <qle/models/defaultprobabilitylatentmodel.hpp>

Inheritance diagram for DefaultLatentModel< copulaPolicy >:

Public Member Functions
	DefaultLatentModel (const std::vector< std::vector< Real >> &factorWeights, LatentModelIntegrationType::LatentModelIntegrationType integralType, const initTraits &ini=initTraits())
	DefaultLatentModel (const Handle< Quote > &mktCorrel, Size nVariables, LatentModelIntegrationType::LatentModelIntegrationType integralType, const initTraits &ini=initTraits())
void	resetBasket (const boost::shared_ptr< Basket > basket) const
Probability	conditionalDefaultProbability (Probability prob, Size iName, const std::vector< Real > &mktFactors) const
Probability	conditionalDefaultProbabilityInvP (Real invCumYProb, Size iName, const std::vector< Real > &m) const
Probability	probOfDefault (Size iName, const Date &d) const
Real	defaultCorrelation (const Date &d, Size iNamei, Size iNamej) const
Probability	probAtLeastNEvents (Size n, const Date &date) const

Protected Member Functions
void	update () override
Probability	conditionalDefaultProbability (const Date &date, Size iName, const std::vector< Real > &mktFactors) const
Probability	condProbProduct (Real invCumYProb1, Real invCumYProb2, Size iName1, Size iName2, const std::vector< Real > &mktFactors) const
Real	conditionalProbAtLeastNEvents (Size n, const Date &date, const std::vector< Real > &mktFactors) const
	Conditional probability of n default events or more.
const boost::shared_ptr< LMIntegration > &	integration () const override
	access to integration:

Protected Attributes
boost::shared_ptr< Basket >	basket_
boost::shared_ptr< LMIntegration >	integration_

Detailed Description

template<class copulaPolicy>
class QuantExt::DefaultLatentModel< copulaPolicy >

Default event Latent Model.

This is a model for joint default events based on a generic Latent Model. It models solely the default events in a portfolio, not making any reference to severities, exposures, etc... An implicit correspondence is stablished between the variables modelled and the names in the basket given by the basket and model variable access indices. The class is parametric on the Latent Model copula.

Constructor & Destructor Documentation

DefaultLatentModel()

DefaultLatentModel	(	const std::vector< std::vector< Real >> &	factorWeights,
		LatentModelIntegrationType::LatentModelIntegrationType	integralType,
		const initTraits &	ini = initTraits()
	)

Parameters

factorWeights	Latent model independent factors weights for each variable.
integralType	Integration type.
ini	Copula initialization if any.

Warning:

Baskets with realized defaults not tested/WIP.

Member Function Documentation

conditionalDefaultProbability() [1/2]

Probability conditionalDefaultProbability	(	Probability	prob,
		Size	iName,
		const std::vector< Real > &	mktFactors
	)		const

Returns the probability of default of a given name conditional on the realization of a given set of values of the model independent factors. The date at which the probability is given is implicit in the probability since theres not other time dependence in this model.

Parameters

prob	Unconditional probability of default.
iName	desired name.
mktFactors	Value of LM independent factors.

Warning:

Most often it is preferred to use the method below avoiding the cumulative inversion.

conditionalDefaultProbabilityInvP()

Probability conditionalDefaultProbabilityInvP	(	Real	invCumYProb,
		Size	iName,
		const std::vector< Real > &	m
	)		const

Returns the probability of default of a given name conditional on the realization of a given set of values of the model independent factors. The date at which the probability is given is implicit in the probability since theres not other time dependent in this model. Same intention as above but provides a performance opportunity, if the integration is along the market factors (as usually is) avoids computing the inverse of the probability on each call.

Parameters

invCumYProb	Inverse cumul of the unconditional probability of default, has to follow the same copula law for results to be coherent
iName	desired name.
m	Value of LM independent factors.

conditionalDefaultProbability() [2/2]

Probability conditionalDefaultProbability ( const Date &  date, Size  iName, const std::vector< Real > &  mktFactors  ) const

protected

Returns the probability of default of a given name conditional on the realization of a given set of values of the model independent factors.

Parameters

date	The date for the probability of default.
iName	desired name.
mktFactors	Value of LM independent factors.

Same intention as the above methods. Usage of this one is typically more expensive because most often the date we call this method with repeats itself and with this one the probability can not be cached outside the call.

condProbProduct()

Probability condProbProduct ( Real  invCumYProb1, Real  invCumYProb2, Size  iName1, Size  iName2, const std::vector< Real > &  mktFactors  ) const

protected

Conditional default probability product, intermediate step in the correlation calculation.

probOfDefault()

Probability probOfDefault	(	Size	iName,
		const Date &	d
	)		const

Computes the unconditional probability of default of a given name. Trivial method for testing

defaultCorrelation()

Real defaultCorrelation	(	const Date &	d,
		Size	iNamei,
		Size	iNamej
	)		const

Pearsons' default probability correlation. Users should consider specialization on the copula type for specific distributions since that might simplify the integrations, most importantly if this is to be used in calibration of observations for factor coefficients as it is expensive to integrate directly.

probAtLeastNEvents()

Probability probAtLeastNEvents	(	Size	n,
		const Date &	date
	)		const

Returns the probaility of having a given or larger number of defaults in the basket portfolio at a given time.