Study of 'value' modelling efficiency in the Monte Carlo method
The paper deals with so called 'value' modelling in the Monte Carlo method and compares its efficiency with that of direct modelling. A terminating homogeneous Markov chain \(x_0, x_1, \ldots , x_N\), specified by the distribution density \(f(x)\) of the initial state \(x_0\) and the substochastic generalized transition density \(k(x',x)\) from state \(x'\) to \(x\), is considered. The value \(\int k(x',x) \, dx = q(x') \leq 1- \delta\) with \(\delta >0\) has the meaning of the trajectory nontermination probability at the given point \(x'\). Let \(\varphi_n(x)\) be the generalised distribution density of the states of number \(n\) (where \(\varphi_0(x) \equiv f(x)\)). Then \(\varphi_n(x)\) is calculated from the distribution in the previous step, e.g., for \(n=1\) \[ \varphi_1(x) = \int f(x') k(x',x) \, dx' = [Kf](x) \;. \] Hence, the generalised distribution density of the mean number of phase states in the chain \(\varphi(x) = \sum_{n=0}^{\infty} \varphi_n(x)\) obeys the integral equation of the second-kind \[ \varphi(x) \int\limits_X k(x',x) \varphi(x') \, dx' + f(x) \;, \] where \(X\) is the \(m\)-dimensional Euclidean space, or \(\varphi = K \varphi + f\). One also considers the adjoint equation \(\varphi^* = K^* \varphi^* + h\), where \([K^* \varphi^*](x) = \int_X k(x,x') \varphi^*(x') \, dx'\). The Monte Carlo methods are generally used to estimate the linear functionals of the form \[ I_h = (\varphi,h) = \int\limits_X \varphi(x) h(x) \, dx = \sum\limits_{n=0}^{\infty} \left(K^n f,h \right) \;. \] If the initial Markov chain is directly modelled, the value \(I_h\) is estimated using the relation \[ I_h = E \left[ \sum\limits_{n=0}^N h \left( x_n \right) \right] \;. \] In the 'value modelling' one uses \[ I_h = E \left[ \sum\limits_{n=0}^N Q_n h \left( x_n \right) \right] \] instead with auxiliary weights \(Q_n\), which are related to transition density of an auxiliary Markov chain. Conditions are obtained under which the value modelling of the initial distribution reduces its variance as compared to direct modelling. The efficiency of the value modelling is analysed for some problems of the transport theory.
- Monte Carlo methods for value-at-risk and conditional value-at-risk: a review
- Efficient method of moments estimation of a stochastic volatility model: A Monte Carlo study
- scientific article; zbMATH DE number 1069624
- Applying the Monte Carlo method for optimum estimation in systems with random structure
- On an efficient multiple time step Monte Carlo simulation of the SABR model
- Comparing stochastic volatility models through Monte Carlo simulations
This page was built for publication: Study of 'value' modelling efficiency in the Monte Carlo method
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5694204)