stat341f11: Difference between revisions

From statwiki
Jump to navigation Jump to search
(First Draft)
Line 1: Line 1:
==[[signupformStat341F11| Editor Sign Up]]==
==[[signupformStat341F11| Editor Sign Up]]==
==Sampling - Sept 20, 2011==
From <math>x ~ f(x)</math> sample <math>x_{1}, x_{2}, ..., x_{1000}</math>
===Sample from uniform distribution.===
Computers cant generate random numbers as they are deterministic but can produce pseudo random numbers.
===Multiplicative Congruential===
*involves three integers
*a, b, m
*an initial value x<sub>0</sub>, we call the seed
*a sequence of integers is defined as
:<math>x_{k+1} = (ax_{k} + b) \mod{m}</math>
example: a=13 b=0 m=31 x<sub>0</sub>=1 creates a uniform histogram
*the first 30 terms in the sequence are a permutation of integers from 1 to 30 then the sequence repeats itself
*values are between 0 and m-1
*if you divide it by m-1 the result is numbers uniformly distributed in the interval of 0 and 1
*in matlab the values choosen are a=7<sup>5</sup> b=0 m=2<sup>31</sup>-1 due to a 1988 paper showing they are optimal
===Inverse Transform Method===
:<math>P(a<x<b)=\int_a^{b} f(x) dx</math>
:<math>cdf=F(x)=P(X<=x)=\int_{-\infty}^{x} f(x) dx</math>
Assume cdf & cdf <sup>-1</sup> can be found
Theorem:
Take u~Uniform[0,1] and let <math>x=F^{-1}(u)</math>. Then x has distribution function <math>F()</math>, where <math>F(x)=P(X<=x)</math>
:Let <math>F^{-1}()</math> denote the inverse of <math>F()</math>
ie <math>F(x)=u \implies x=F^{-1}(u)</math>
Take the the exponential distribution for example
:<math>f(x)={\lambda}e^{-{\lambda}x}</math>
:<math>F(x)=\int_0^x {\lambda}e^{-{\lambda}u} du</math>
:<math>F(x)=1-{\lambda}e^{-{\lambda}x}</math>
:<math>{\lambda}e^{-{\lambda}x}=1-F(x)</math>
:<math>{-{\lambda}x}=ln(1-F(x))</math>
:<math>x=\frac{ln(1-F(x))}{-\lambda}</math>
:<math>F^{-1}(x)=\frac{-ln(1-F(x))}{\lambda}</math>
1. Draw u~Uniform[0,1]
2. <math>x=\frac{-ln(1-F(u))}{\lambda}
<math>p(F^{-1}(u)<=x)</math>
<math>p(F(F^{-1}(u))<=F(x))</math>
<math>p(u<=F(x))</math>
<math>=F(x)</math>
=> data points generated have cdf F(x)
This same technique can be applied to the discrete case
Generate a discrete random variable x that has probability mass function <math>p(x=x_i)=P_i </math> where <math>x_0<x_1<x_2... and \sum p_i=1</math>
1 u~uniform[0,1]
2 x=x<sub>i</sub> if <math>F(x_{i-1})<u<=F(x_i)</math

Revision as of 19:02, 21 September 2011

Editor Sign Up

Sampling - Sept 20, 2011

From [math]\displaystyle{ x ~ f(x) }[/math] sample [math]\displaystyle{ x_{1}, x_{2}, ..., x_{1000} }[/math]

Sample from uniform distribution.

Computers cant generate random numbers as they are deterministic but can produce pseudo random numbers.

Multiplicative Congruential

  • involves three integers
  • a, b, m
  • an initial value x0, we call the seed
  • a sequence of integers is defined as
[math]\displaystyle{ x_{k+1} = (ax_{k} + b) \mod{m} }[/math]

example: a=13 b=0 m=31 x0=1 creates a uniform histogram

  • the first 30 terms in the sequence are a permutation of integers from 1 to 30 then the sequence repeats itself
  • values are between 0 and m-1
  • if you divide it by m-1 the result is numbers uniformly distributed in the interval of 0 and 1
  • in matlab the values choosen are a=75 b=0 m=231-1 due to a 1988 paper showing they are optimal

Inverse Transform Method

[math]\displaystyle{ P(a\lt x\lt b)=\int_a^{b} f(x) dx }[/math]
[math]\displaystyle{ cdf=F(x)=P(X\lt =x)=\int_{-\infty}^{x} f(x) dx }[/math]

Assume cdf & cdf -1 can be found

Theorem: Take u~Uniform[0,1] and let [math]\displaystyle{ x=F^{-1}(u) }[/math]. Then x has distribution function [math]\displaystyle{ F() }[/math], where [math]\displaystyle{ F(x)=P(X\lt =x) }[/math]

Let [math]\displaystyle{ F^{-1}() }[/math] denote the inverse of [math]\displaystyle{ F() }[/math]

ie [math]\displaystyle{ F(x)=u \implies x=F^{-1}(u) }[/math] Take the the exponential distribution for example

[math]\displaystyle{ f(x)={\lambda}e^{-{\lambda}x} }[/math]
[math]\displaystyle{ F(x)=\int_0^x {\lambda}e^{-{\lambda}u} du }[/math]
[math]\displaystyle{ F(x)=1-{\lambda}e^{-{\lambda}x} }[/math]
[math]\displaystyle{ {\lambda}e^{-{\lambda}x}=1-F(x) }[/math]
[math]\displaystyle{ {-{\lambda}x}=ln(1-F(x)) }[/math]
[math]\displaystyle{ x=\frac{ln(1-F(x))}{-\lambda} }[/math]
[math]\displaystyle{ F^{-1}(x)=\frac{-ln(1-F(x))}{\lambda} }[/math]

1. Draw u~Uniform[0,1] 2. [math]\displaystyle{ x=\frac{-ln(1-F(u))}{\lambda} \lt math\gt p(F^{-1}(u)\lt =x) }[/math] [math]\displaystyle{ p(F(F^{-1}(u))\lt =F(x)) }[/math] [math]\displaystyle{ p(u\lt =F(x)) }[/math] [math]\displaystyle{ =F(x) }[/math] => data points generated have cdf F(x) This same technique can be applied to the discrete case Generate a discrete random variable x that has probability mass function [math]\displaystyle{ p(x=x_i)=P_i }[/math] where [math]\displaystyle{ x_0\lt x_1\lt x_2... and \sum p_i=1 }[/math] 1 u~uniform[0,1] 2 x=xi if <math>F(x_{i-1})<u<=F(x_i)</math