Regression Estimators Related to Multinormal
Distributions: Computer Experiences in Root Finding
Istvan Decik[l]
[1] Operations Research Group, Dept. of Differential Equations
Technical University of Budapest
H-l1l1 Budapest, XI. Muegyetem rkp. 3.
email: deak@inf.bme.hu
Several simple regression estimators can be constructed to approximate the dis-
tribution function of the m-dimensional normal distribution along a line. These
functions can be used to find the border points of the feasible region of probability
constrained stochastic programming models. Computer experiences show a fast
and robust behaviour of the root finding techniques.
Keywords: multinormal distribution, stochastic programming, regression
estimators, quantile computation
1 Introduction
Consider the m-dimensional normal distribution with expected value 0 and
correlation matrix R. Its distribution function and density function are given
as
4'(h) (1)
¢(z)
Computation of the function values 4'(h) is required in numerical optimiza-
tion procedures of stochastic programming problems, when the random variables
of the model have a joint normal distribution. This is the case in solution pro-
cedures of the STABIL stochastic programming model [11] and the two-stage
model [9]. Other problems, where computation of (1) is required can be found
in diverse areas of statistics and engineering ([1], [6], [4]).
K. Marti et al. (eds.), Stochastic Programming Methods and Technical Applications
© Springer-Verlag Berlin Heidelberg 1998