10.06.2016 Views

eldo_user

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

How to Specify a Distribution Function<br />

Monte Carlo Analysis<br />

How to Specify a Distribution Function<br />

By definition, “simple” random variables refer to statistical distributions that can easily be<br />

generated on a computer. The context usually dictates which random variables are meant. For<br />

example, the uniform distribution U([0, 1]) is simple, and so are the normal or <strong>user</strong>-defined<br />

cumulative distributions in most circumstances. In other terms, the simple distributions are<br />

elementary building blocks, and “complex” distributions can be defined as the result of complex<br />

transformations of the other distributions. The simple random variables are often associated to<br />

independent variables such as some process parameters or environmental conditions. In some<br />

cases, the PDK models are derived from Principal Components Analysis, and the principal<br />

components are simple Gaussian distributions.<br />

The available distributions are grouped into two classes. The first class is associated to<br />

parametric models. These models aim to describe probability distributions of a random variable<br />

with the aid of a limited number of parameters Θ. The probability density of X can be expressed<br />

as f X (x, Θ). The list of these distributions is given here:<br />

• Uniform Distribution<br />

• Gaussian or Normal Distribution<br />

• Weighted Uniform and Gaussian Distributions<br />

The second class represents distributions that are specified by a non-parametric procedure. A<br />

typical example is given by the so-called weighted uniform or gaussian that is defined implicitly<br />

by an accept/reject algorithm. The list of these distributions is given here:<br />

• User-defined Distribution<br />

• NOR/UNI Distributions<br />

• Weighted Uniform and Gaussian Distributions<br />

Note<br />

Some of these distributions, such as the “weighted” UNIF/GAUSS, may seem unusual or<br />

non-standard for many <strong>user</strong>s. They represent the legacy of past experimentations in Eldo<br />

and other third part tools. The implementation of those distributions are described.<br />

Uniform Distribution<br />

The uniform distribution defines equal probability over a given range for a continuous<br />

distribution. The parameters are Θ = (L, U), with the constraint L < U. The probability density<br />

function is expressed as:<br />

Eldo® User's Manual, 15.3 431

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!