Simulation of multivariate normal rectangle probabilities and their derivatives. Theoretical and computational results (Q1915466): Difference between revisions
From MaRDI portal
Added link to MaRDI item. |
Removed claims |
||
Property / author | |||
Property / author: Vassilis Argyrou Hajivassiliou / rank | |||
Property / author | |||
Property / author: Paul A. Ruud / rank | |||
Revision as of 14:20, 11 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Simulation of multivariate normal rectangle probabilities and their derivatives. Theoretical and computational results |
scientific article |
Statements
Simulation of multivariate normal rectangle probabilities and their derivatives. Theoretical and computational results (English)
0 references
6 November 1996
0 references
An extensive literature in econometrics and in numerical analysis has considered the problem of evaluating the multiple integral \[ P({\mathbf B}; \mu, \Omega) = \int^b_a n(v - \mu, \Omega) dv \equiv {\mathbf E}_v \mathbf{1} (V \in {\mathbf B}), \] where \(V\) is an \(m\)-dimensional normal vector with mean \(\mu\), covariance matrix \(\Omega\), and density \(n(v - \mu, \Omega)\), and \(\mathbf{1} (V \in {\mathbf B})\) is an indicator for the event \({\mathbf B} = \{V \mid a < V < b)\). A leading case of such an integral is the negative orthant probability, where \({\mathbf B} = \{V \mid V < 0)\). The problem is computationally difficult except in very special cases. The multinomial probit model used in econometrics and biometrics has cell probabilities that are negative orthant probabilities, with \(\mu\) and \(\Omega\) depending on unknown parameters (and, in general, on covariates). Estimation of this model requires, for each trial parameter vector and each observation in a sample, evaluation of \(P({\mathbf B}; \mu, \Omega)\) and of its derivatives with respect to \(\mu\) and \(\Omega\). This paper surveys Monte Carlo techniques that have been developed for approximations of \(P({\mathbf B}; \mu, \Omega)\) and its linear and logarithmic derivatives, that limit computation while possessing properties that facilitate their use in iterative calculations for statistical inference: the Crude Frequency Simulator, Normal Importance Sampling, a kernel-smoothed frequency simulator, etc. We also discuss Gauss and FORTRAN implementations of these algorithms and present our computational experience with them.
0 references
simulation estimation
0 references
Monte Carlo integration
0 references
discrete choice models
0 references
importance sampling
0 references
acceptance/rejection
0 references
Gibbs resampling
0 references
crude frequency simulator
0 references
kernel-smoothed frequency simulator
0 references
multiple integral
0 references
negative orthant probability
0 references
multinomial probit model
0 references
derivatives
0 references
FORTRAN
0 references