Transcript Chapter 2. Random Variables

Chapter 2. Random Variables
2.1
2.2
2.3
2.4
2.5
2.6
NIPRL
Discrete Random Variables
Continuous Random Variables
The Expectation of a Random Variable
The Variance of a Random Variable
Jointly Distributed Random Variables
Combinations and Functions of Random Variables
2.1 Discrete Random Variable
2.1.1 Definition of a Random Variable (1/2)
• Random variable
– A numerical value to each outcome of a particular
experiment
S
R
-3
NIPRL
-2
-1
0
1
2
3
2.1.1 Definition of a Random Variable (2/2)
• Example 1 : Machine Breakdowns
– Sample space : S  {electrical , m echanical , m isuse}
– Each of these failures may be associated with a repair cost
– State space : {50, 200, 350}
– Cost is a random variable : 50, 200, and 350
NIPRL
2.1.2 Probability Mass Function (1/2)
• Probability Mass Function (p.m.f.)
– A set of probability value p i assigned to each of the values
taken by the discrete random variable x i
– 0  p i  1 and  i p i  1
– Probability : P ( X  x i )  p i
NIPRL
2.1.2 Probability Mass Function (1/2)
• Example 1 : Machine Breakdowns
– P (cost=50)=0.3, P (cost=200)=0.2,
P (cost=350)=0.5
– 0.3 + 0.2 + 0.5 =1
f (x)
xi
50
200
350
pi
0.3
0.2
0.5
0.5
0.3
0.2
50
NIPRL
200
350
C ost($)
2.1.3 Cumulative Distribution Function (1/2)
• Cumulative Distribution Function
– Function : F ( x )  P ( X  x ) F ( x )
– Abbreviation : c.d.f


P( X  y)
y: y  x
F (x)
1.0
0.5
0.3
0
NIPRL
50
200
350
x ($cost)
2.1.3 Cumulative Distribution Function (2/2)
• Example 1 : Machine Breakdowns
  x  50  F ( x )  P (cost  x )  0
50  x  200  F ( x )  P (cost  x )  0.3
200  x  350  F ( x )  P (cost  x )  0.3  0.2  0.5
350  x    F ( x )  P (cost  x )  0.3  0.2  0.5  1.0
NIPRL
2.2 Continuous Random Variables
2.2.1 Example of Continuous Random Variables (1/1)
• Example 14 : Metal Cylinder Production
– Suppose that the random variable X is the diameter of a
randomly chosen cylinder manufactured by the company.
Since this random variable can take any value between
49.5 and 50.5, it is a continuous random variable.
NIPRL
2.2.2 Probability Density Function (1/4)
• Probability Density Function (p.d.f.)
– Probabilistic properties of a continuous random variable
f ( x)  0

NIPRL
f ( x ) dx  1
statespace
2.2.2 Probability Density Function (2/4)
• Example 14
– Suppose that the diameter of a metal cylinder has a p.d.f
f ( x )  1.5  6( x  50.2)
2
for 49.5  x  50.5
f ( x )  0, elsew here
f (x)
49.5
NIPRL
50.5
x
2.2.2 Probability Density Function (3/4)
• This is a valid p.d.f.

50.5
49.5
(1 .5  6 ( x  5 0 .0 ) ) d x  [1 .5 x  2 ( x  5 0 .0 ) ] 49.5
2
3 50.5
 [1 .5  5 0 .5  2 (5 0 .5  5 0 .0 ) ]
3
 [1 .5  4 9 .5  2 (4 9 .5  5 0 .0 ) ]
3
 7 5 .5  7 4 .5  1 .0
NIPRL
2.2.2 Probability Density Function (4/4)
• The probability that a metal cylinder has a diameter between
49.8 and 50.1 mm can be calculated to be

5 0 .1
4 9 .8
(1 .5  6 ( x  5 0 .0 ) ) d x  [1 .5 x  2 ( x  5 0 .0 ) ] 4 9 .8
2
3
5 0 .1
 [1 .5  5 0 .1  2 (5 0 .1  5 0 .0 ) ]
3
 [1 .5  4 9 .8  2 ( 4 9 .8  5 0 .0 ) ]
3
 7 5 .1 4 8  7 4 .7 1 6  0 .4 3 2
f (x)
49.5
NIPRL
49.8
50.1
50.5
x
2.2.3 Cumulative Distribution Function (1/3)
• Cumulative Distribution Function
 F ( x)  P ( X  x) 
 f ( x) 

x

f ( y )dy
dF ( x)
dx
 P (a  X  b)  P ( X  b)  P ( X  a )
 F (b )  F ( a )
 P (a  X  b)  P (a  X  b)
NIPRL
2.2.2 Probability Density Function (2/3)
• Example 14
F ( x)  P ( X  x) 

x
(1.5  6( y  50.0) ) dy
2
49.5
 [1.5 y  2( y  50.0) ] 49.5
3 x
 [1.5 x  2( x  50.0) ]  [1.5  49.5  2(49.5  50.0) ]
3
3
 1.5 x  2( x  50.0)  74.5
3
P (49.7  X  50.0)  F (50.0)  F (49.7 )
 (1.5  50.0  2(50.0  50.0)  74.5)
3
 (1.5  49.7  2(49.7  50.0)  74.5)
3
 0.5  0.104  0.396
NIPRL
2.2.2 Probability Density Function (3/3)
P (49.7  X  50.0)  0.396
1
P ( X  50.0)  0.5
F (x)
P ( X  49.7)  0.104
49.5
NIPRL
49.7
50.0
50.5
x
2.3 The Expectation of a Random Variable
2.3.1 Expectations of Discrete Random Variables (1/2)
• Expectation of a discrete random variable with p.m.f
P ( X  xi )  p i

E(X ) 
p i xi
i
• Expectation of a continuous random variable with p.d.f f(x)
E(X ) 

xf ( x ) d x
state sp ace
• The expected value of a random variable is also called the
mean of the random variable
NIPRL
2.3.1 Expectations of Discrete Random Variables (2/2)
• Example 1 (discrete random variable)
– The expected repair cost is
E (cost )  ($50  0.3)  ($200  0.2)  ($350  0.5)  $230
NIPRL
2.3.2 Expectations of Continuous Random Variables (1/2)
• Example 14 (continuous random variable)
– The expected diameter of a metal cylinder is
E(X ) 

5 0 .5
x (1 .5  6 ( x  5 0 .0 ) ) d x
2
4 9 .5
– Change of variable: y=x-50
E (x) 


0.5
 0.5
0.5
 0.5
( y  50)(1.5  6 y ) dy
2
(  6 y  300 y  1.5 y  75) dy
3
2
 [  3 y / 2  100 y  0.75 y  75 y ]  0.5
4
3
2
 [25.09375]  [  24.90625]  50.0
NIPRL
0.5
2.3.2 Expectations of Continuous Random Variables (2/2)
• Symmetric Random Variables
– If x has a p.d.f f ( x ) that is
symmetric about a point 
so that
f (  x)  f (  x)
– Then, E ( X )  
E(X )  
(why?)
– So that the expectation of
the random variable is equal
to the point of symmetry
NIPRL
f (x)

x
E(X ) 
 xf ( x ) d x



-

xf ( x ) d x +  xf ( x ) d x

 y  2  x



-
 
NIPRL

xf ( x ) d x +  

-
yf ( y ) d y
2.3.3 Medians of Random Variables (1/2)
• Median
– Information about the “middle” value of the random variable
F ( x )  0.5
• Symmetric Random Variable
– If a continuous random variable is symmetric about a point  ,
then both the median and the expectation of the random
variable are equal to 
NIPRL
2.3.3 Medians of Random Variables (2/2)
• Example 14
F ( x )  1.5 x  2( x  50.0)  74.5  0.5
3
x  5 0 .0
NIPRL
2.4 The variance of a Random Variable
2.4.1 Definition and Interpretation of Variance (1/2)
• Variance(  )
– A positive quantity that measures the spread of the
distribution of the random variable about its mean value
– Larger values of the variance indicate that the distribution is
more spread out
2
– Definition:
V ar ( X )  E (( X  E ( X )) )
2
 E ( X )  ( E ( X ))
2
• Standard Deviation
– The positive square root of the variance
– Denoted by 
NIPRL
2
2.4.1 Definition and Interpretation of Variance (2/2)
V a r ( X )  E (( X  E ( X )) )
2
 E(X
2
 2 X E ( X )  ( E ( X )) )
 E(X
2
)  2 E ( X ) E ( X )  ( E ( X ))
 E(X
2
)  ( E ( X ))
2
2
2
f (x)
Two distribution with
identical mean values but
different variances
x
NIPRL
2.4.2 Examples of Variance Calculations (1/1)
• Example 1
V ar ( X )  E (( X  E ( X )) )   p i ( x i  E ( X ))
2
2
i
 0 .3(5 0  2 3 0 )  0 .2 (2 0 0  2 3 0 )  0 .5(3 5 0  2 3 0 )
2
 1 7 ,1 0 0  
 
NIPRL
2
17,100  130.77
2
2
2.4.3 Chebyshev’s Inequality (1/1)
• Chebyshev’s Inequality
– If a random variable has a mean  and a variance  2, then
P (   c  X    c  )  1 
1
c
2
for c  1
– For example, taking c  2 gives
P (   2  X    2 )  1 
1
2
NIPRL
2
 0.75
• Proof


2



( x   ) f ( x ) dx 
2

( x   ) f ( x ) dx  c 
2
| x   |  c
 P (| x   | c )  1 / c
2
2

| x   |  c
2
 P (| x   | c )  1  P (| x   | c )  1  1 / c
NIPRL
f ( x ) dx .
2
2.4.4 Quantiles of Random Variables (1/2)
• Quantiles of Random variables
– The p th quantile of a random variable X
F ( x)  p
– A probability of p that the random variable takes a value
less than the p th quantile
• Upper quartile
– The 75th percentile of the distribution
• Lower quartile
– The 25th percentile of the distribution
• Interquartile range
– The distance between the two quartiles
NIPRL
2.4.4 Quantiles of Random Variables (2/2)
• Example 14
3
F ( x )  1.5 x  2( x  50.0)  74.5 for 49.5  x  50.5
– Upper quartile : F ( x )  0.75
x  5 0 .1 7
– Lower quartile : F ( x )  0.25
x  4 9 .8 3
– Interquartile range : 5 0 .1 7  4 9 .8 3  0 .3 4
NIPRL
2.5 Jointly Distributed Random Variables
2.5.1 Jointly Distributed Random Variables (1/4)
• Joint Probability Distributions
– Discrete
P ( X  x i , Y  y j )  p ij  0
s a tis fyin g

i
p ij  1
j
– Continuous
f ( x , y )  0 satisfying

state space
NIPRL
f ( x , y ) dxdx  1
2.5.1 Jointly Distributed Random Variables (2/4)
• Joint Cumulative Distribution Function
– Discrete
F ( x , y )  P ( X  xi , Y  y j )
– Continuous
F ( x, y ) 
 
p ij
i : xi  x j : y j  y
F ( x, y ) 
NIPRL

x
w  

y
z  
f ( w , z ) d zd w
2.5.1 Jointly Distributed Random Variables (3/4)
• Example 19 : Air Conditioner Maintenance
– A company that services air conditioner units in residences
and office blocks is interested in how to schedule its
technicians in the most efficient manner
– The random variable X, taking the values 1,2,3 and 4, is the
service time in hours
– The random variable Y, taking the values 1,2 and 3, is the
number of air conditioner units
NIPRL
2.5.1 Jointly Distributed Random Variables (4/4)
Y=
number
of units
• Joint p.m.f
X=service time

1
2
3
4
1
0.12
0.08
0.07
0.05
2
0.08
0.15
0.21
0.13
i
p ij  0.12  0.18
j

 0.07  1.00
• Joint cumulative
distribution function
F (2, 2)  p11  p12  p 21  p 22
 0.12  0.18  0.08  0.15
3
NIPRL
0.01
0.01
0.02
0.07
 0.43
2.5.2 Marginal Probability Distributions (1/2)
• Marginal probability distribution
– Obtained by summing or integrating the joint probability
distribution over the values of the other random variable
– Discrete
P ( X  i )  pi 

p ij
j
– Continuous
f X (x) 
NIPRL



f ( x, y )dy
2.5.2 Marginal Probability Distributions (2/2)
• Example 19
– Marginal p.m.f of X
3
P ( X  1) 

p1 j  0 .1 2  0 .0 8  0 .0 1  0 .2 1
j 1
– Marginal p.m.f of Y
4
P (Y  1) 

i 1
NIPRL
p i1  0.12  0.08  0.07  0.05  0.32
• Example 20: (a jointly continuous case)
• Joint pdf:
f ( x, y )
• Marginal pdf’s of X and Y:
NIPRL
f X ( x) 

f ( x , y ) dy
fY ( y ) 

f ( x , y ) dx
2.5.3 Conditional Probability Distributions (1/2)
• Conditional probability distributions
– The probabilistic properties of the random variable X under
the knowledge provided by the value of Y
– Discrete
p ij
P ( X  i, Y  j )
p i| j  P ( X  i | Y  j ) 

P (Y  j )
p j
– Continuous
f X |Y  y ( x ) 
f ( x, y )
fY ( y )
– The conditional probability distribution is a probability
distribution.
NIPRL
2.5.3 Conditional Probability Distributions (2/2)
• Example 19
– Marginal probability distribution of Y
P (Y  3)  p  3  0.01  0.01  0.02  0.07  0.11
– Conditional distribution of X
p1|Y  3  P ( X  1 | Y  3) 
NIPRL
p13
p 3

0.01
0.11
 0.091
2.5.4 Independence and Covariance (1/5)
• Two random variables X and Y are said to be independent if
– Discrete
p ij  p i  p  j
for all values i of  X and  j of Y
– Continuous
f ( x , y )  f X ( x ) fY ( y )
fo r a ll x a n d y
– How is this independency different from the independence
among events?
NIPRL
2.5.4 Independence and Covariance (2/5)
• Covariance
C ov ( X , Y )  E (( X  E ( X ))(Y  E (Y )))
 E ( X Y )  E ( X ) E (Y )
C ov( X , Y )  E (( X  E ( X ))(Y  E (Y )))
 E ( XY  XE (Y )  E ( X )Y  E ( X ) E (Y ))
 E ( XY )  E ( X ) E (Y )  E ( X ) E (Y )  E ( X ) E (Y )
 E ( XY )  E ( X ) E (Y )
– May take any positive or negative numbers.
– Independent random variables have a covariance of zero
– What if the covariance is zero?
NIPRL
2.5.4 Independence and Covariance (3/5)
• Example 19 (Air conditioner maintenance)
E ( X )  2.59,
4
E ( XY ) 
3
  ijp
i 1
E (Y )  1.79
ij
j 1
 (1  1  0.12)  (1  2  0.08)

 (4  3  0.07 )  4.86
C ov ( X , Y )  E ( X Y )  E ( X ) E (Y )
 4.86  (2.59  1.79)  0.224
NIPRL
2.5.4 Independence and Covariance (4/5)
• Correlation:
C orr ( X , Y ) 
C ov ( X , Y )
V ar ( X )V ar (Y )
– Values between -1 and 1, and independent random
variables have a correlation of zero
NIPRL
2.5.4 Independence and Covariance (5/5)
• Example 19: (Air conditioner maintenance)
V ar ( X )  1.162, V ar (Y )  0.384
C orr ( X , Y ) 
C ov ( X , Y )
V ar ( X )V ar (Y )

NIPRL
0.224
1.162  0.384
 0.34
• What if random variable X and Y have linear relationship, that is,
Y  aX  b
where a  0
C ov ( X , Y )  E [ X Y ]  E [ X ] E [Y ]
 E [ X ( aX  b )]  E [ X ] E [ aX  b ]
 aE [ X ]  bE [ X ]  aE [ X ]  bE [ X ]
2
2
 a ( E [ X ]  E [ X ])  aV ar ( X )
2
C orr ( X , Y ) 
2
C ov ( X , Y )
V ar ( X )V ar (Y )
aV ar ( X )

That is, Corr(X,Y)=1 if a>0; -1 if a<0.
NIPRL
2
V ar ( X ) a V ar ( X )
2.6 Combinations and Functions of Random
Variables
2.6.1 Linear Functions of Random Variables (1/4)
• Linear Functions of a Random Variable
– If X is a random variable and Y  a X  b
for some numbers a , b  R then E (Y )  aE ( X )  b
and V ar (Y )  a 2 V ar ( X )
• Standardization
-If a random variable X has an expectation of  and a
2
variance of  ,
X 
1
  
Y 


X  

  
has an expectation of zero and a variance of one.
NIPRL
2.6.1 Linear Functions of Random Variables (2/4)
• Example 21:Test Score Standardization
– Suppose that the raw score X from a particular testing
procedure are distributed between -5 and 20 with an
expected value of 10 and a variance 7. In order to
standardize the scores so that they lie between 0 and 100,
the linear transformation Y  4 X  2 0 is applied to the
scores.
NIPRL
2.6.1 Linear Functions of Random Variables (3/4)
– For example, x=12 corresponds to a standardized score of
y=(4ⅹ12)+20=68
E (Y )  4 E ( X )  20  (4  10)  20  60
V ar(Y )  4 V ar( X )  4  7  112
2
Y 
NIPRL
112  10.58
2
2.6.1 Linear Functions of Random Variables (4/4)
• Sums of Random Variables
– If X 1 and X 2 are two random variables, then
E ( X 1  X 2 )  E ( X 1 )  E ( X 2 )            ( w hy ?)
and
V ar ( X 1  X 2 )  V ar ( X 1 )  V ar ( X 2 )  2C ov ( X 1 , X 2 )
– If X 1 and X 2 are independent, so that C o v ( X 1 , X 2 )  0
then
V ar ( X 1  X 2 )  V ar ( X 1 )  V ar ( X 2 )
NIPRL
• Properties of C ov ( X , X )
1
2
C ov ( X 1 , X 2 )  E [ X 1 X 2 ]  E [ X 1 ] E [ X 2 ]
 C ov ( X 1 , X 2 )  C ov ( X 2 , X 1 )
Cov ( X 1 , X 1 )  Var ( X 1 ) C ov ( X 2 , X 2 )  Var ( X 2 )
C ov ( X 1  X 2 , X 1 )  C ov ( X 1 , X 1 )  C ov ( X 2 , X 1 )
C ov ( X 1  X 2 , X 1  X 2 )  C ov ( X 1 , X 1 )  C ov ( X 1 , X 2 ) 
C ov ( X 2 , X 1 )  C ov ( X 2 , X 2 )
 Var ( X 1  X 2 )  Var ( X 1 )  Var ( X 2 )  2 Cov ( X 1 , X 2 )
NIPRL
2.6.2 Linear Combinations of Random Variables (1/5)
• Linear Combinations of Random Variables
– If X 1 , , X n is a sequence of random variables and a1 ,
and b are constants, then
E ( a1 X 1 
 a n X n  b )  a1 E ( X 1 ) 
 an E ( X n )  b
– If, in addition, the random variables are independent, then
V ar ( a1 X 1 
NIPRL
2
 a n X n  b )  a1 V ar ( X 1 ) 
2
 a n V ar ( X n )
, an
2.6.2 Linear Combinations of Random Variables (2/5)
• Averaging Independent Random Variables
– Suppose that X 1 , , X n is a sequence of independent
random variables with an expectation  and a variance  2 .
– Let
– Then
X 
X1 
 Xn
n
E(X )  
and
V ar ( X ) 

2
n
– What happened to the variance?
NIPRL
2.6.2 Linear Combinations of Random Variables (3/5)
1
E(X )  E  X1 
n
1
1
 

1

n

Xn
n

1
n
E ( X 1) 

1
n
E(X n)
n
1
V ar ( X )  V ar  X 1 
n
2
1
  
n
NIPRL

2
 1
 X n     V ar ( X 1 ) 
n
 n
1
2
2

1
  
n
2


2
n
2
1
  V ar ( X n )
n
2.6.2 Linear Combinations of Random Variables (4/5)
• Example 21
– The standardized scores of the two tests are
Y1 
10
3
X1
Y2 
and
5
3
X2 
50
3
– The final score is
Z 
2
3
NIPRL
Y1 
1
3
Y2 
20
9
X1 
5
9
X2 
50
9
2.6.2 Linear Combinations of Random Variables (5/5)
– The expected value of the final score is
E (Z )

20
9
E ( X1) 
5
9
E(X 2) 
50
9
 20
 5
 50
 
 18     30  
9
 9
 9

 6 2 .2 2
– The variance of the final score is
5
50 
 20
V ar ( Z )  V ar 
X1  X 2 

9
9 
 9
2
2
 20 
5


V
ar
(
X
)

1

   V ar ( X 2 )
 9 
9
2
2
 20 
5


24


   60  137.04
9


9
NIPRL
2.6.3 Nonlinear Functions of a Random Variable (1/3)
• Nonlinear function of a Random Variable
– A nonlinear function of a random variable X is another
random variable Y=g(X) for some nonlinear function g.
Y  X , Y 
2
X , Y  e
X
– There are no general results that relate the expectation and
variance of the random variable Y to the expectation and
variance of the random variable X
NIPRL
2.6.3 Nonlinear Functions of a Random Variable (2/3)
– For example,
f(x)=1
E(x)=0.5
f X ( x )  1 for 0  x  1
f ( x)  0
elsew here
0
F X ( x )  x fo r 0  x  1
1
x
f(y)=1/y
E(y)=1.718
– Consider
Y e
X
w here 1  Y  2.718
– What is the pdf of Y?
NIPRL
1.0
2.718
y
2.6.3 Nonlinear Functions of a Random Variable (3/3)
• CDF methd
FY ( y )  P (Y  y )  P ( e
X
 y )  P ( X  ln( y ))  F x (ln( y ))  ln( y )
– The p.d.f of Y is obtained by differentiation to be
fY ( y ) 
dFY ( y )

dy
1
for 1  y  2.718
y
• Notice that
E (Y ) 

E (Y )  e
NIPRL
2.718
z 1
E(X )
zf Y ( z ) dz 
e
0.5

2.718
z 1
 1.649
1dz  2.718  1  1.718
• Determining the p.d.f. of nonlinear relationship between r.v.s:
Given f X ( x ) and Y  g ( X ) ,what is f Y ( y ) ?
If
x1 , x 2 ,
, x n are all its real roots, that is,
y  g ( x1 )  g ( x 2 ) 
then
n
fY ( y ) 
f X ( xi )
 | g '( x ) |
i 1
where
g '( x ) 
dg ( x )
dx
NIPRL
i
 g ( xn )
•
Example: determine f Y ( y )
f X ( x )  1 for 0  x  1
f ( x)  0
elsew here
y  g ( x)  e
 ln y  x
dg
x
dx
fY ( y ) 
NIPRL
|y|

X
w here 1  Y  2.718
x
-> one root is possible in 0  x  1
e  y
1
Y e
1
y