Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.
-
Upload
alexandra-knight -
Category
Documents
-
view
217 -
download
0
Transcript of Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.
![Page 1: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/1.jpg)
Hamid R. RabieeFall 2009
Stochastic Processes
Review of Elementary ProbabilityLecture I
![Page 2: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/2.jpg)
Outline
History/Philosophy Random Variables Density/Distribution Functions Joint/Conditional Distributions Correlation Important Theorems
![Page 3: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/3.jpg)
History & Philosophy
Started by gamblers’ dispute Probability as a game analyzer ! Formulated by B. Pascal and P.
Fermet First Problem (1654) :
“Double Six” during 24 throws
First Book (1657) : Christian Huygens, “De Ratiociniis in
Ludo Aleae”, In German, 1657.
![Page 4: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/4.jpg)
History & Philosophy (Cont’d)
Rapid development during 18th Century
Major Contributions: J. Bernoulli (1654-1705) A. De Moivre (1667-1754)
![Page 5: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/5.jpg)
History & Philosophy (Cont’d)
A renaissance: Generalizing the concepts from mathematical analysis of games to analyzing scientific and practical problems: P. Laplace (1749-1827)
New approach first book: P. Laplace, “Théorie Analytique des
Probabilités”, In France, 1812.
![Page 6: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/6.jpg)
History & Philosophy (Cont’d)
19th century’s developments: Theory of errors Actuarial mathematics Statistical mechanics
Other giants in the field: Chebyshev, Markov and Kolmogorov
![Page 7: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/7.jpg)
History & Philosophy (Cont’d)
Modern theory of probability (20th) : A. Kolmogorov : Axiomatic approach
First modern book: A. Kolmogorov, “Foundations of
Probability Theory”, Chelsea, New York, 1950
Nowadays, Probability theory as a part of a theory called Measure theory !
![Page 8: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/8.jpg)
History & Philosophy (Cont’d)
Two major philosophies: Frequentist Philosophy
Observation is enough Bayesian Philosophy
Observation is NOT enough Prior knowledge is essential
Both are useful
![Page 9: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/9.jpg)
History & Philosophy (Cont’d)
Frequentist philosophy There exist fixed
parameters like mean,. There is an underlying
distribution from which samples are drawn
Likelihood functions(L()) maximize parameter/data
For Gaussian distribution the L() for the mean happens to be 1/Nixi or the average.
Bayesian philosophy Parameters are variable Variation of the
parameter defined by the prior probability
This is combined with sample data p(X/) to update the posterior distribution p(/X).
Mean of the posterior, p(/X),can be considered a point estimate of .
![Page 10: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/10.jpg)
History & Philosophy (Cont’d)
An Example: A coin is tossed 1000 times, yielding 800 heads and
200 tails. Let p = P(heads) be the bias of the coin. What is p?
Bayesian Analysis Our prior knowledge (believe) : (Uniform(0,1)) Our posterior knowledge :
Frequentist Analysis Answer is an estimator such that
Mean : Confidence Interval :
1p 200800 1 ppnObservatiop
p̂
8.0ˆ pE 95.0826.0ˆ774.0 pP
![Page 11: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/11.jpg)
History & Philosophy (Cont’d)
Further reading: http://www.leidenuniv.nl/fsw/verduin/s
tathist/stathist.htm http://www.mrs.umn.edu/~sungurea/i
ntrostat/history/indexhistory.shtml www.cs.ucl.ac.uk/staff/D.Wischik/Talks
/histprob.pdf
![Page 12: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/12.jpg)
Outline
History/Philosophy Random Variables Density/Distribution Functions Joint/Conditional Distributions Correlation Important Theorems
![Page 13: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/13.jpg)
Random Variables Probability Space
A triple of represents a nonempty set, whose
elements are sometimes known as outcomes or states of nature
represents a set, whose elements are called events. The events are subsets of . should be a “Borel Field”.
represents the probability measure.
Fact:
PF ,,
F
F
P
1P
![Page 14: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/14.jpg)
Random Variables (Cont’d)
Random variable is a “function” (“mapping”) from a set of possible outcomes of the experiment to an interval of real (complex) numbers.
In other words :
rX
IFX
RI
F
:
:Outcomes
Real Line
![Page 15: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/15.jpg)
Random Variables (Cont’d)
Example I : Mapping faces of a dice to the first six
natural numbers. Example II :
Mapping height of a man to the real interval (0,3] (meter or something else).
Example III : Mapping success in an exam to the
discrete interval [0,20] by quantum 0.1 .
![Page 16: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/16.jpg)
Random Variables (Cont’d)
Random Variables Discrete
Dice, Coin, Grade of a course, etc. Continuous
Temperature, Humidity, Length, etc.
Random Variables Real Complex
![Page 17: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/17.jpg)
Outline
History/Philosophy Random Variables Density/Distribution Functions Joint/Conditional Distributions Correlation Important Theorems
![Page 18: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/18.jpg)
Density/Distribution Functions
Probability Mass Function (PMF) Discrete random variables Summation of impulses The magnitude of each impulse
represents the probability of occurrence of the outcome
Example I: Rolling a fair dice 1 2 3 4 5 6
61
X
XP
6
16
1
i
iXPMF
![Page 19: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/19.jpg)
Density/Distribution Functions (Cont’d)
Example II: Summation of two fair dices
Note : Summation of all probabilities should be equal to ONE. (Why?)
2 73 4 5 6
61
X
XP
8 9 10 11 12
![Page 20: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/20.jpg)
Density/Distribution Functions (Cont’d)
Probability Density Function (PDF) Continuous random variables The probability of occurrence of
will be
2,
20dx
xdx
xx
X
XP
x
xP
dxxP .
![Page 21: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/21.jpg)
Density/Distribution Functions (Cont’d)
Some famous masses and densities Uniform Density
Gaussian (Normal) Density
beginUendUa
xf .1
a1
X
XP
a
2.1
X
XP
,2.
1 2
2
.2 Nexfx
![Page 22: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/22.jpg)
Density/Distribution Functions (Cont’d)
Binomial Density
Poisson Density
!1:
1
xxxNote
xexf
x
nNn ppn
Nnf
.1.
0 N.pn
nf
N
x
xf
Important Fact: !
..1.: .
n
pNepp
n
NNlargelySufficientFor
npNnnN
![Page 23: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/23.jpg)
Density/Distribution Functions (Cont’d)
Cauchy Density
Weibull Density
22
1
xxf
X
XP
k
xk
exk
xf
1
![Page 24: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/24.jpg)
Density/Distribution Functions (Cont’d)
Exponential Density
Rayleigh Density
00
0...
x
xexUexf
xx
2
2 2
2
.
x
exxf
![Page 25: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/25.jpg)
Density/Distribution Functions (Cont’d)
Expected Value The most likelihood value
Linear Operator
Function of a random variable Expectation
dxxfxXE X.
dxxfxgXgE X.
bXEabXaE ..
![Page 26: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/26.jpg)
Density/Distribution Functions (Cont’d)
PDF of a function of random variables Assume RV “Y” such that The inverse equation may have more
than one solution called PDF of “Y” can be obtained from PDF of “X” as
follows
XgY
YgX 1
nXXX ,...,, 21
n
i
xx
iXY
i
xgdxd
xfyf
1
![Page 27: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/27.jpg)
Density/Distribution Functions (Cont’d)
Cumulative Distribution Function (CDF) Both Continuous and Discrete Could be defined as the integration of PDF
x
XX
X
dxxfxF
xXPxFxCDF
.
X
XPDF
x
CDF(x)
![Page 28: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/28.jpg)
Density/Distribution Functions (Cont’d)
Some CDF properties Non-decreasing Right Continuous F(-infinity) = 0 F(infinity) = 1
![Page 29: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/29.jpg)
Outline
History/Philosophy Random Variables Density/Distribution Functions Joint/Conditional Distributions Correlation Important Theorems
![Page 30: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/30.jpg)
Joint/Conditional Distributions Joint Probability Functions
Density Distribution
Example I In a rolling fair dice experiment represent
the outcome as a 3-bit digital number “xyz”.
..0
1;161
0;131
1;031
0;061
,,
WO
yx
yx
yx
yx
yxf YX
1106
0113
0102
xyz
1015
1004
0011
x y
YX
YX
dydxyxf
yYandxXPyxF
,
,
,
,
![Page 31: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/31.jpg)
Joint/Conditional Distributions (Cont’d)
Example II Two normal random variables
What is “r” ?
Independent Events (Strong Axiom)
yx
yx
y
y
x
x yxryx
r
yx
YX er
yxf
.
2
12
1
2,
2
2
2
2
2
1...2
1,
yfxfyxf YXYX .,,
![Page 32: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/32.jpg)
Joint/Conditional Distributions (Cont’d)
Obtaining one variable densitydensity functions
DistributionDistribution functions can be obtained just from the density functions. (How?)
dxyxfyf
dyyxfxf
YXY
YXX
,
,
,
,
![Page 33: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/33.jpg)
Joint/Conditional Distributions (Cont’d)
Conditional Density Function Probability of occurrence of an event if
another event is observed (we know what “Y” is).
Bayes’ Rule
yf
yxfyxf
Y
YXYX
,,
dxxfxyf
xfxyfyxf
XXY
XXYYX
.
.
![Page 34: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/34.jpg)
Joint/Conditional Distributions (Cont’d)
Example I Rolling a fair dice
X : the outcome is an even number Y : the outcome is a prime number
Example II Joint normal (Gaussian) random variables
3
1
2161,
YP
YXPYXP
2
212
1
21..2
1 y
y
x
x yr
x
r
x
YX er
yxf
![Page 35: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/35.jpg)
Joint/Conditional Distributions (Cont’d)
Conditional Distribution Function
Note that “y” is a constant during the integration.
dtytf
dtytf
dxyxf
yYwhilexXPyxF
YX
x
YX
x
YX
YX
,
,
,
,
![Page 36: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/36.jpg)
Joint/Conditional Distributions (Cont’d)
Independent Random Variables
Remember! Independency is NOT heuristic.
xf
yf
yfxf
yf
yxfyxf
X
Y
YX
Y
YXYX
.
,,
![Page 37: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/37.jpg)
Joint/Conditional Distributions (Cont’d)
PDF of a functions of joint random variables Assume that The inverse equation set has a
set of solutions Define Jacobean matrix as follows
The joint PDF will be
YXgVU ,),(
VUgYX ,),( 1
nn YXYXYX ,,...,,,, 2211
VY
UX
VX
UXJ
n
i yxyx
iiYXVU
iiJtdeterminanabsolute
yxfvuf
1 ,,
,,
.
,,
![Page 38: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/38.jpg)
Outline
History/Philosophy Random Variables Density/Distribution Functions Joint/Conditional Distributions Correlation Important Theorems
![Page 39: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/39.jpg)
Correlation Knowing about a random variable “X”,
how much information will we gain about the other random variable “Y” ?
Shows linear similarity
More formal:
Covariance is normalized correlation
YXEYXCrr .,
YXYX YXEYXEYXCov ...),(
![Page 40: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/40.jpg)
Correlation (cont’d)
Variance Covariance of a random variable with itself
Relation between correlation and covariance
Standard Deviation Square root of variance
22XX XEXVar
222XXXE
![Page 41: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/41.jpg)
Correlation (cont’d)
Moments nth order moment of a random variable “X”
is the expected value of “Xn”
Normalized form
Mean is first moment Variance is second moment added by
the square of the mean
nXn XEM
nn XEM
![Page 42: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/42.jpg)
Outline
History/Philosophy Random Variables Density/Distribution Functions Joint/Conditional Distributions Correlation Important Theorems
![Page 43: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/43.jpg)
Important Theorems Central limit theorem
Suppose i.i.d. (Independent Identically Distributed) RVs “Xk” with finite variances
Let
PDF of “Sn” converges to a normal distribution as nn increases, regardless to the density of RVs.
Exception : Cauchy Distribution (Why?)
n
innn XaS
1
.
![Page 44: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/44.jpg)
Important Theorems (cont’d)
Law of Large Numbers (Weak) For i.i.d. RVs “Xk”
0Prlim 10
X
n
ii
n n
X
![Page 45: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/45.jpg)
Important Theorems (cont’d)
Law of Large Numbers (Strong) For i.i.d. RVs “Xk”
Why this definition is stronger than before?
1limPr 1
X
n
ii
n n
X
![Page 46: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/46.jpg)
Important Theorems (cont’d)
Chebyshev’s Inequality Let “X” be a nonnegative RV Let “c” be a positive number
Another form:
It could be rewritten for negative RVs. (How?)
XEc
cX1
Pr
2
2
Pr X
XX
![Page 47: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/47.jpg)
Important Theorems (cont’d)
Schwarz Inequality For two RVs “X” and “Y” with finite second
moments
Equality holds in case of linear dependency.
222 E.E.E YXYX
![Page 48: Hamid R. Rabiee Fall 2009 Stochastic Processes Review of Elementary Probability Lecture I.](https://reader034.fdocuments.us/reader034/viewer/2022051820/56649dbe5503460f94ab1dd1/html5/thumbnails/48.jpg)
Next Lecture
Elements of Stochastic Elements of Stochastic Processes Processes