Scribd
Upload
38 views

Modification Box Muller

This document discusses methods for generating random variables from a standard normal (N(0,1)) distribution. It explains that only an N(0,1) distribution needs to be generated since other normal distributions can be obtained through linear transformations. It then describes the Box-Muller, polar, and crude central limit theorem methods. The Box-Muller method generates two independent standard normal random variables from two uniformly distributed random numbers. The polar method also generates standard normal random variables from uniformly distributed random numbers by transforming them to a polar coordinate system.

Uploaded by

Promach
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
38 views

Modification Box Muller

This document discusses methods for generating random variables from a standard normal (N(0,1)) distribution. It explains that only an N(0,1) distribution needs to be generated since other normal distributions can be obtained through linear transformations. It then describes the Box-Muller, polar, and crude central limit theorem methods. The Box-Muller method generates two independent standard normal random variables from two uniformly distributed random numbers. The polar method also generates standard normal random variables from uniformly distributed random numbers by transforming them to a polar coordinate system.

Uploaded by

Promach
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Normal Random Variables

Special Properties

Need only generate N(0, 1)


If X is IID N(0, 1), then Y = X+ is N(,)

Given X ~ N(0,1), can generate


X ~ N(, 2) as X = + X
So sufficient to be able to generate X
Inversion not possible
Acceptance/rejection an option
But better methods have been developed

1) Crude (Central Limit Theorem) Method


U U(0, 1), hence E(U) = .5 and V(U) = 1/12.

Set Y =

Ui, hence E(Y) = n/2 and V(Y) = n/12

i=1

By CLT, (Y-n/2)/(n/12)1/2 D N(0, 1) as n +


Consider

12

Ui-6 (i.e., n = 12).


i=1

However n may be too small!

Box-Muller Method
Generate U1, U2 U(0,1) random numbers
Generate X 1 = 2 ln U1 cos ( 2 U 2 )
and
X 2 = 2 ln U1 sin ( 2 U 2 )
X1 and X2 will be iid N(0,1) random variables

Box-Muller Explanation
If X1 and X2 iid N(0,1)
D2 = X12 + X22 is 22 which is the SAME as
an Exponential(2)
D = 2 ln U
X1 = D cos
X2 = D sin
D
X2
where = 2U2
X1

Polar Method
1.
2.
3.
4.
5.

Polar Method Graphically


(-1,1)

Let U1, U2 be U(0,1),


define Vi = 2Ui 1: Vi is U(-1,1)
Define S = V12 + V22
If S>1, go to 1.
2 ln S
If S1, then Y =

(1,1)

V1

V2

1. X1=V1Y and X2=V2Y are iid N(0,1)

(-1,-1)

(1,-1)

Lets Look at some Normal Random Variates...

Notes
X12 + X 22 is distributed 22 or exponential (2)

or ( V12 + V22)Y2 = -2ln(S)


is true if
Theorem: S ~ U(0, 1) given S 1.

You might also like