Professional Documents
Culture Documents
SMA 2231 Probability and Statistics III
SMA 2231 Probability and Statistics III
SMA 2231 Probability and Statistics III
COURSE OUTLINE
4. Bivariate moment generating function (MGF) and change of variable techniques for
bivariate distribution.
5. Stochastic independence.
8. Independence of sample mean and Variance for normal mean and variance for normal
distribution.
References:
Sheaffer.
Discrete Bivariate Probability Distribution
realizable value of Y. let the probability that X takes values x and Y takes values y be
denoted by
P ( X = x, Y = y )
i) f ( x, y ) ≥ 0
ii) ∑∑ f (x, y ) = 1
x y
Example
Consider an experiment of tossing a pair of dice. The sample space contains 36 sample points
corresponding to 6 x 6 =36 ways which the numbers may appear on the faces of the die
2nd dice
1 2 3 4 5 6
-2-
Any of the following random variables could be defined over the sample space and might be of
The 36 sample points associated with the experiment are equiprobable and correspond to the
36 numerical events ( x1 , x 2 ) . Thus if ones are obtained in the throw of the dice, the simple
event is (1, 1) . Throw a 2 on die 1 and a 3 on die 2 would be the sample event (2, 3) . Because all
pairs ( x1 , x 2 ) occur with the same relative frequency a probability of 1/36 would be assigned to
For this example, the intersection ( x1 , x 2 ) contains only one sample point. Hence the bivariate
probability function is
⎧1
⎪ x1 = 1, 2, 3, 4 ,5, 6 x 2 = 1, 2, 3, 4 ,5, 6
P(x1 , x 2 ) = ⎨ 36
⎪⎩0 Otherwise
Definition:
Let the random variable X1 , X2 take a countable number of pairs of real values ( X 1 , X 2 ) .If
a) P( X1 , X 2 ) ≥ 0
b) ∑∑ P( X , X
x1 x2
1 2 ) = 1 and
b d
c) For any constants a, b, c and d , P (a ≤ X 1 ≤ b , and , c ≤ X 2 ≤ d ) = ∑ ∑ P( X 1 , X2)
x1 = a x 2 = c
-3-
Then X1 and X2 are said to have a joint (or bivariate) discrete probability distribution with joint
Example:
Using the results of tossing of two dice, calculate the P(2 ≤ X 1 ≤ 3,1 ≤ X 2 ≤ 2)
Solution:
4 1
= =
36 9
Definition:
For any random variables X1 and X2, the joint (bivariate) distribution function, F (a, b ) is
given by
F (a, b ) = P ( X 1 ≤ a, X 2 ≤ b )
F (a, b ) = P( X 1 ≤ a, X 2 ≤ b ) = ∑ ∑ P(x x ) 1 2
all x1 all x2
Example
F (2, 3) = P(X 1 ≤ 2 , X 2 ≤ 3)
1 1 1 1 1 1
= + + + + +
36 36 36 36 36 36
1
=
36
Example
Suppose the random variables X1 and X2 have the following probability distribution.
-4-
X1
0 1 2
0 1 2 1
X2 9 9 9
1 2 2 0
9 9
2 1 0 0
9
Find
Solution
a) F (-1, 2) = P(X 1 ≤ -1 , X 2 ≤ 2) = P( φ ) = 0
=8/9
F (5,7) = P(X 1 ≤ 5 , X 2 ≤ 7) = 1
Definition:
The random variables X1, X2 are said to have a bivariate (joint) continuous distribution with
-5-
2. f ( x1 , x 2 ) is continuous except along a countable number of points and curves.
∞ ∞
3. ∫ ∫ f (x , x )dx dx
− ∞− ∞
1 2 1 2 =1
b d
4. For constants a, b, c and d, P(a ≤ X 1 ≤ b, and , c ≤ X 2 ≤ d ) = ∫ ∫ f ( x1 , x2 )dx1 dx2
a c
Definition
x1 x 2
F( x1 , x2 ) = ∫ ∫ f (t , t
- ∞- ∞
1 2 )dt2 dt1
For any real numbers x1 and x 2 , then X1 and X2 are said to be jointly continuous random
Example:
Suppose that a radioactive particle is randomly located in a square with sides of unit length.
That is, if two regions of equal areas are considered, the particle is equally likely to be in
either. Let X1 and X2 denote the co ordinates locating the particles. A reasonable model for the
relative frequency histogram for X1 and X2 would be the bivariate analogues to the univariate
uniform distribution.
⎧1 ,0 ≤ x1 ≤ 1 , 0 ≤ x 2 ≤ 1
f ( x1 , x 2 ) = ⎨
⎩0 ,Otherwise
-6-
Solution
i) Diagram
0.40.2
ii) F (0.2, 0.4 ) = ∫ ∫ f (x , x )dx dx 1 2 1 2
-∞ -∞
0.40.2
= ∫ ∫ f (1)dx dx
- ∞ -∞
1 2
0 .4
∫ [x ]
0 .4
= 1 0 dx 2
-∞
0 .4
= ∫ 0 . 2 dx
-∞
2
= [0.2 x2 ]0
0.4
= 0.08
Note: The probability F (0.2, 0.4) corresponds to the volume under f (x1 , x 2 ) = 1 over the
0.50.3
= ∫ ∫ f ((x , x ) )dx dx
0 0.1
1 2 1 2
0.50.3
= ∫ ∫ dx dx
0 0.1
1 2
= 0.1
theorem
-7-
Theorem
Let X1 and X2 be random variables discrete or continuous with the joint distribution
function f ( x1 , x 2 ) , then
a) F (− ∞,−∞ ) = F (− ∞, X 2 ) = F ( X 1 ,−∞ ) = 0
b) F (∞, ∞ ) = 1
Example
⎧3x ,0 ≤ x 2 ≤ x1 ≤ 1
f ( x1 , x 2 ) = ⎨ 1
⎩0 ,Otherwise
Solution
0 .5 x1
= ∫ ∫ 3x
0 . 25 0 . 25
1 dx 2 dx 1
0.5
= ∫ 3 x [x ]
x1
1 2 0.25 dx1
0.25
0.5
1
= ∫
0 .25
3 x1 [ x1 − ]dx1
4
0 .5
⎡ 3 3 2⎤
= ⎢ x1 − x1 ⎥
⎣ 8 ⎦ 0.25
⎡ 1 3 ⎛ 1 ⎞⎤ ⎡ 1 3 ⎛ 1 ⎞⎤
= ⎢ − ⎜ ⎟⎥ ⎢ − ⎜ ⎟⎥
⎣ 8 8 ⎝ 4 ⎠ ⎦ ⎣ 64 8 ⎝ 16 ⎠ ⎦
-8-
5
=
128
Question
Three fair coins are tossed independently one of the variables of interest is X1 = the umber of
heads. Let X2 denote the amount of money won on a side set in the following manner. If the
first head occurs on the first toss, you win $ 1. If the first head occurs on toss 2 or on toss 3 you
b) What is the probability that less than three heads occur and you win $ 1 or less? (I.e.
F (2, 1)
Solution
X1
a)
0 1 2 3
-1 1 0 0 0
8
1 0 1 2 1
8 8 8
X2
2 0 1 1 0
8 8
3 0 1 0 0
8
i)
ii) F (2 ,1) = 1
2
Question
-9-
⎧k (1 - x 2 ) ,0 ≤ x1 ≤ x 2 ≤ 1
f (x1 , x 2 ) = ⎨
⎩0 ,Otherwise
1 x2
∫ ∫ k (1 − x
0 0
2 )dx 1 dx 2 = 1
Question
Let X1 and X2 denote the proportions of time, out of one working day, that employee A and B,
respectively, actually spend performing their assigned tasks. The joint relative frequency
⎧ x + x2 ,0 ≤ x1 ≤ 1;0 ≤ x 2 ≤ 1
f (x1 , x 2 ) = ⎨ 1
⎩0 ,elsewhere
Question
⎧6xy 2 ,0 ≤ x ≤ 1 , 0≤ y ≤1
f ( x, y ) = ⎨
⎩0 , elsewhere
Solution
- 10 -
∞ ∞
i) ∫ ∫ f (x , x )dx dx
− ∞− ∞
1 2 1 2 =1
∫ [2 xy ]
0.5 0.75 0.5
3 0.75
= ∫ ∫ 6 xy dydx =
2
0.5 dx
0 0.5 0
0.5 0.5
= ∫ 2 x [ 27 ] [19 ]
64 − 8 dx = ∫ 2 x 64 dx
1
0 0
= x2[ 19
0.5
64 0
] = 19
64 ⋅ 14 = 19
256
Question
i) f ( x, y ) = ⎨
(
⎧k x 2 + 2 y ) 0 < x < 1; 1 < y < 3
⎩0 elsewhere
(
⎧k x + e 2 x
ii) f ( x, y ) = ⎨
) 0 < x < 2; 0 < y < 1
⎩0 elsewhere
Question
⎧3
⎪ x( y + x ) 0 < x < 1; 0 < y < 2
f ( x, y ) = ⎨ 5
⎪⎩0 elsewhere
Question
Find F (x, y )
- 11 -
Question
( )(
⎧ 1 − e−x 1 − e−y
f ( x, y ) = ⎨
)
x > 0; y > 0
⎩0 elsewhere
Find the density of ( x, y ) and P[1 < x < 2 and 3 < y < 5 ]
Solutions
Definition
Let X1 and X2 be jointly discrete random variables with probability function P (x1 , x 2 ) . Then
Similarly, If X1 and X2 are jointly continuous random variables with joint density
function f (x1 , x 2 ) , then the marginal density function of X1 and X2, respectively are given by
∞ ∞
Example
⎧a( y + 3x + 1) , x = 0, 1, 2; y = 1, 3
f ( x, y ) = ⎨
⎩0 ,elsewhere
- 12 -
Solution
0 1 2 P1(y)
1 2a 5a 8a 15a
Y
3 4a 7a 10a 21a
36a = 1
1
a=
36
6 1 ⎫
f (0 ) = 6a = =
36 6 ⎪
⎪ ⎧ 6( x + 1) (x + 1)
12 1 ⎪ ⎪ = for x = 0, 1, 2
f (1) = 12a = = ⎬ f ( x ) = ⎨ 36 6
36 3 ⎪ ⎪⎩0 elsewhere
18 1 ⎪
f (2 ) = 18a = =
36 2 ⎪⎭
15 5 ⎫
f (1) = 15a = =
36 12 ⎪⎪ ⎧
⎬ f (y) = ⎨
21 7 ⎪ ⎩ checkk
f (3) = 21a = =
36 12 ⎪⎭
E (x ) =
4
3
( )
E x2 =
7
3
Var ( x ) = E ( x ) − (E ( x ))
2 2
7 16 5 35
Var (x ) = − = Check Var ( y ) =
3 9 9 36
- 13 -
Example
X1
0 1 2 P2(x2)
a) X1
b) X2
Solution
I.
X1 0 1 2
II.
X2 0 1 2
Example
- 14 -
X
1 2 3 P1(y)
4 0 1/3 0 1/3
1 1
i) Find the marginal values of PMFs F2 (4), F1 (2 ) [ F2 (4 ) = F1 (2 ) = ]
3 2
Example
⎧x + y
⎪ , for x = 1, 2, 3; y = 1, 2
f (x, y ) = ⎨ 21
⎪⎩0 , elsewhere
3y + 6 y + 2 2x + 3
Answer: [ f 2 ( y ) = = f1 (x ) =
21 7 21
Example
⎧ λ x + y e −2 λ
⎪ x = 0, 1, 2; y = 0, 1, 2
f ( x, y ) = ⎨ x! y!
⎪0
⎩ elsewhere
- 15 -
Example
⎧2x ,0 ≤ x1 ≤ 1;0 ≤ x 2 ≤ 1
f (x1 , x 2 ) = ⎨ 1
⎩0 ,elsewhere
Solution
∞
f 1 ( x1 ) =
-∞
∫ f( x x 1 2 )dx 2
1
= ∫2x
0
1 dx 2
= [x1 x 2 ]0 = 2x1
1
⎧2 x ,0 ≤ x1 ≤ 1
⇒ f 1 ( x1 ) = ⎨ 1
⎩0 , elsewhere
Similarly
∞
f 2 ( x2 ) = ∫ f( x x
-∞
1 2 )dx1
1
= ∫2x
0
1 dx 1
=x1
2 1
0
] =1
⎧1 ,0 ≤ x 2 ≤ 1
⇒ f 2 ( x2 ) = ⎨
⎩0 ,elsewhere
- 16 -
Conditional Distribution Function
Definition
Suppose X1 and X2 are jointly discrete random variables with probability function
P (x1 x 2 ) = P ( X 1 = x1 X 2 = x 2 )
P( X 1 = x1 X 2 = x 2 )
=
P( X 2 = x 2 )
P(x1 , x 2 ) f (x1 , x 2 )
= = Provided P2 ( x 2 ) > 0
P( x 2 ) f (x 2 )
P (x1 , x 2 ) f ( x1 , x 2 )
P (x 2 x1 ) = = Provided P ( x1 ) > 0
P (x1 ) f ( x1 )
Special cases
f ( x, y ) f ( x ) f ( y )
f (x y ) = = = f (x ) .
f (y) f (y)
Similarly
f ( x, y ) f ( x ) f ( y )
f (y x) = = = f (y)
f (x ) f (x )
- 17 -
Example
0 1 2 P2(x2)
2 1/15 0 0 1/15
a) Find the conditional distribution of X1 given that (i). X2=1, (ii). X2=2
Solution
P(x1 , x2 )
(
a) (i). P x1 x2 =) where x2 =1
P2 (x2 )
P( x1 , x2 = 1)
P(x1 x2 = 1) =
P2 ( x2 = 1)
P(x1 ,1)
=
P2 (1)
2
P(0,1) 15 1
= =
P2 (1) 8 4
15
6
P(1,1) 15 3
= =
P2 (1) 8 4
15
- 18 -
0
P(2,1) 15
= =0
P2 (1) 8
15
X1 0 1 2
b) Complete
Definition
Let X1 and X2 be jointly continuous random variables with joint density f (x1 , x 2 ) and
X 2 = x 2 is given by
⎧ f ( x1 , x2 )
⎪ , f 2 (x2 ) > 0
f1 (x1 x2 ) = ⎨ f 2 ( x2 )
⎪0
⎩ ,elsewhere
⎧ f ( x1 , x 2 )
⎪ , f1 (x1 ) > 0
f 2 (x 2 x1 ) = ⎨ f 1 ( x1 )
⎪0
⎩ ,elsewhere
Example
⎧1
⎪ ,0 ≤ x1 ≤ x2 , 0 ≤ x2 ≤ 2
f ( x2 , x1 ) = ⎨ 2
⎪⎩0 ,elsewhere
Find
- 19 -
Solution
∞
f 2 (x 2 ) = ∫ f (x x )dx 1 2 1
-∞
1
1 1
= ∫
0
2
dx 1 =
2
x2
⎧1
⎪ x , x1 ≤ x2 ≤ 2
i.e. f 2 (x 2 ) = ⎨ 2 2
⎪⎩0 ,elsewhere
f ( x1 , x 2 )
By definition, f (x1 x 2 ) =
f 2 (x 2 )
1
2 1
=
1
(x2 ) x2
2
⎧1
⎪ , 0 ≤ x1 ≤ x 2 ≤ 2
f 1 ( x1 x 2 ) = ⎨ x 2
⎪0, , elsewhere
⎩
Now
0.5
P ( X 1 ≤ 0.5 X 2 = 1) = ∫ f (x 1 x 2 = 1)dx1
-∞
0.5
1 1
∫ 1 dx
-∞
1 =
2
Example
In a group of nine executives of a certain business firm, four are married, three have never
married and two are divorced. Three of the executives are to be selected for promotion. Let X1
denote the number of married executives and X2 the number of never married executives
- 20 -
among the three selected for promotion. Assuming that the three are randomly selected from
b) Find the marginal probability distribution of X1, the number of married executives
d) Let X3 denote the number of divorced executives among the three selected for
Solution
N = 9, n = 3, r1 = 4, r2 = 3, r3 = 2 , r = r1 + r2 + r3 = N
⎧ ⎛ 4 ⎞⎛ 3 ⎞⎛ 2 ⎞
⎪ ⎜⎜ ⎟⎟⎜⎜ ⎟⎟⎜⎜ ⎟⎟
⎪⎪ ⎝ x1 ⎠⎝ x 2 ⎠ ⎝ 3 − x1 − x 2 ⎠ ⎧0 ≤ x1 ≤ 3
⎪
P ( x1 , x 2 ) = ⎨ ⎛9⎞ ⎨0 ≤ x 2 ≤ 3, and
⎪ ⎜⎜ ⎟⎟ ⎪0 ≤ x + x ≤ 3
⎪ ⎝3⎠ ⎩ 1 2
⎪⎩0, elsewhere
3
b) P1 ( x1 ) = ∑ P(x , x ) = ∑ P(x , x )
1 2 1 2
all x2 x =0
Then
X1 0 1 2 3
P1 (x1 ) 5 20 15 2
42 42 42 42
P( x1 = 1, x 2 = 2 )
c) (i) P1 (x1 = 1 X 2 = 2 ) =
P2 (x 2 = 2 )
- 21 -
= P2 ( x2 = 2 ) = ∑ p ( X ,2 )
1
all x1
2
=
3
ii. = P (x 2 = 2 X 1 = 1) = ?
P( X 3 = 1, X 2 = 1)
a. P( X 3 = 1 X 2 = 1) =
P2 ( X 2 = 1)
Note that. X 1 = 1, X 2 = 1, X 3 = 1
⎛ 4 ⎞⎛ 3 ⎞⎛ 2 ⎞
⎜⎜ ⎟⎟⎜⎜ ⎟⎟⎜⎜ ⎟⎟
1 1 1
P(x3 = 1, X 2 = 1) = ⎝ ⎠⎝ ⎠⎝ ⎠
⎛9⎞
⎜⎜ ⎟⎟
⎝3⎠
P2 (x 2 = 1) = ∑ P( X 1 ,1) 8
allx1 =
15
=?
Question
Solution
f (x , y )
f (y x) =
f (x )
∞ ∞
f (x ) = ∫ f (x , y ) dy = ∫ e
y
dy
−∞ x
[
= − e −y ] ∞
x
= e−x
- 22 -
The conditional PDF for y given x is
f (x , y ) e − y
f (y x) = = − x = e x − y for y > x ,0 < x < y < ∞
f (x ) e
Question
Solution
f (x , y )
f (y x) =
f (x )
[ ]
1
f ( x ) = ∫ ( x + y ) dy = xy + 12 y 2
1
0
= x+ 1
2
0
f (y x) =
(x + y ) = 2(x + y )
x + 12 2x + 1
2( x + y )
y
F (y x) = ∫ dy
0
2x + 1
=
2
2x + 1
[
xy + 12 y 2 ]y
0
=
2
2x + 1
( )
xy + 12 y 2 or
2 xy + y 2
2x + 1
x y
Note F ( x , y ) = P[X ≤ x ,Y ≤ y ] = ∫ ∫ f (x , y ) dydx
− ∞− ∞
- 23 -
Distribution of continuous variables
Definition
If X1 and X2 are jointly continuous random variables with joint density function f ( X 2 , X 1 )
f ( x1 , x 2 )
x1
The random variables X1 and X2 have the following joint probability density function
a)
Solution
f ( x1 , x 2 )
f (x 2 x1 ) =
f ( x1 )
Now
1
f ( x1 ) = ∫ f ( x1 , x2 )dx2
0
1
= ∫ (x1 + x2 )dx2
0
- 24 -
1
⎡ x2 2 ⎤
= ⎢x x + ⎥
1 2
⎣ 2 ⎦ 0
⎛ 1⎞
= ⎜ x1 + ⎟-0
⎝ 2⎠
Now
f ( x1 , x 2 )
f (x 2 x1 ) =
f 1 (x1 )
⎧ ( x1 + x 2 )
⎪ ,0 < x2 < 1
= ⎨ (x1 + 0.5)
⎪0
⎩ ,otherwise
0 .5
0 . 25 + x 2
= ∫ 0 .25 + 0 .5 dx
0
2
0 .5
⎡ 1 x2 ⎤
2
= ⎢ (0 .25 x 2 + )⎥
⎣⎢ 0 .75 2 ⎦⎥ 0
1
=
3
Question
If X1 is the total time between a customer’s arrival in the store and leaving the service window
and X2 is the time spent in line before reaching the window, then the joint density of these
variables is given as
⎧e -x1 ,0 ≤ x 2 ≤ 1 < ∞
f (x1 , x 2 ) = ⎨
⎩0 ,elsewhere
- 25 -
c) Find P ( x 1 − x 2 ) ≥ 1 (Note the X 1 − X 2 is the time spent at the service window).
Answer [ e-1]
d) If 2 minutes elapse between a customer’s arrival at the store and his departure from the
service window, find the probability that he waited in line less than one minute to reach
Definition
X2 have joint distribution function f ( x1 , x 2 ) then X1 and X2 are said to be independent iff
Note
1. If X1 and X2 are Discrete random variables with joint probability function P(x1 , x 2 ) and
independent iff
f ( x1 , x 2 ) = f1 ( x1 ) f 2 ( x 2 )
- 26 -
Question
⎧4x x ,0 ≤ x1 ≤ 1; 0 ≤ x 2 ≤ 1
f ( x1 , x 2 ) = ⎨ 1 2
⎩0 , elsewhere
Solution
1
f1 ( x1 ) = ∫ f (x1 , x2 )dx2
0
1
= ∫ 4x
0
1 x 2 dx 2
1
⎡ xx 2 ⎤
= ⎢4 1 2 )⎥ = 2 x1 ,0 ≤ x1 ≤ 1
⎣ 2 ⎦ 0
Similarly,
1
f 2 (x2 ) = ∫ f ( x1, x2 )dx1
0
= 2x2 ,0 ≤ x2 ≤1
independent.
Question
⎧2 ,0 ≤ x 2 ≤ x1 ; 0 ≤ x1 ≤ 1
Let f (x1 , x 2 ) = ⎨
⎩0 , elsewhere
- 27 -
Question
Solution
Question
⎧x + y ,0 ≤ x ≤ 1, 0 ≤ y ≤ 1
f ( x, y ) = ⎨
⎩0 , elsewhere
Solution
Trinomial distribution
Definition
Discrete random variables X1, X2 are said to have the trinomial distribution with positive
parameters n , P1 + P2 such that n is an integer and 0 < P1 + P2 < 1 , if the joint probability
⎧
P1 1 P2 2 (1 - p1 - p2 ) 1 2 , x1 , x2 are non- negativeintegersand that x1 + x2 ≤ n
n! x x n- x - x
⎪
P(x1 , x2 ) = ⎨ x1 ! x2 ! (n - x1 - x2 )!
⎪0
⎩ ,elsewhere
Note:
- 28 -
Question
distribution.
Solution
n− x2
P2 (x2 ) = ∑ P1 1 P2 2 (1 - p1 - p2 ) 1 2
n! x x n- x - x
x1 =0 x1 ! x2 ! (n - x1 - x2 )!
n− x2
(n - x2 )!
P1 1 (1 - p1 - p2 ) 1 2
n!
= P2 2 ∑
x x n- x - x
x2 ! (n - x2 )! x1 =0 x1 ! (n - x1 - x2 )!
P2 2 (1 - p1 - p2 + p1 ) 2
n!
=
x n- x
x2 ! (n - x2 )!
P2 2 (1 - p2 ) 2
n!
=
x n- x
x2 ! (n - x2 )!
Now
P(x1 , x2 )
P(x1 / x2 ) =
P2 (x2 )
P2 2 P1 1 (1 - p1 - p2 ) 1 2
n! x x n- x - x
x ! x ! (n - x1 - x2 )!
= 1 2
P2 2 (1 - p2 ) 2
n! x n- x
x2 ! (n - x2 )!
(n - x2 )! P1 x (1 - p1 - p 2 )n- x - x
1 1 2
=
x1 ! (n - x1 - x 2 )! (1 - p 2 )n- x 2
x n-x1 -x2
⎛n- x2 ⎞⎛ P1 ⎞ ⎛1- p2 - p1 ⎞
1
= ⎜⎜ ⎟⎟⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟
⎝x1 ⎠⎝1- P2 ⎠ ⎝ 1- p2 ⎠
P1
This is a binomial distribution with parameters n - x2 and
1 - P2
- 29 -
Note
P1 1 (1- p1 ) 1
n!
P1 (x1 ) =
x n- x
x1 ! (n - x1 )!
Example
A bag contains three white, two black and four red marbles. Four marbles are drawn at random
with replacement; calculate the probability that the sample contains just one white marble
Example
⎧
⎪ x2 −2λ
⎪ λ e ⎧x = 0,1,2,3,4...x2
P(x1 , x2 ) = ⎨ ,⎨ 1
⎪ x1!(x2 − x1 )! ⎩x2 = 0,1,2,3,....
⎪⎩0, otherwise
Find the marginal distribution of X1 and X2 and the conditional distribution of X1 given X2
Solution
∞
λx e−2λ
2
P1 ( x1 ) = ∑ x !(x
x2 = x1 − x1 )!
1 2
λ x e −2 λ
1 ∞
λx 2 − x1
=
x1 !
∑ (x
x 2 = x1 − x1 )!
2
λ x e −2 λ
1
= eλ
x1!
- 30 -
λx e −λ
1
= x 1 = 0 ,1 , 2 ....
x1 !
x2
λ x e −2 λ
2
P2 ( x2 ) = ∑
x1 = 0 x1!( x2 − x1 )!
λ x e −2 λ
2 x2
x2!
=
x2!
∑
x1 = 0 x1 ! ( x 2 − x1 )!
1 x11 x 2 − x1
But
x2
⎛ x2 ⎞ x1 x2 − x1
∑ ⎜⎜ x ⎟⎟1 1 = (1 + 1) 2 = 2 x2
x
x1 = 0 ⎝ 1 ⎠
Thus
P2 ( x 2 ) =
(2λ )
x 2
e −2 λ
, x 2 = 0,1,2,3....
x2 !
P( x1 , x2 )
P(x1 x2 ) =
P2 ( x2 )
λ x e −2 λ
2
x 1 ! ( x 2 − x 1 )!
=
(2 λ )x 2 e − 2 λ
x2!
x2
⎛1⎞ x2 !
=⎜ ⎟
⎝2⎠ x1!( x 2 − x1 )!
⎛ x2 ⎞
x2
⎛1⎞
=⎜ ⎟ ⎜⎜ ⎟⎟
⎝2⎠ ⎝ x1 ⎠
- 31 -
Bivariate Expectations
Let the random variables X, Y have joint probability mass function/ density function
E [g ( x , y )] = ∑ ∑ g ( x , y ) f ( x , y ) Discrete case
all x all y
∞ ∞
E [g ( x , y )] = ∫ ∫ g (x , y ) f (x , y ) dydx Continuous case
− ∞− ∞
CASE 1
g (x , y ) = (x − μ x ) (y − μ y ) where μ x = E [X ] and μ y = E [Y ]
[
E [g ( x , y )] = E ( x − μ x ) ( y − μ y ) ]
= ∑∑ ( x − μ x ) ( y − μ y ) f ( x , y ) (Covariance of X, Y)
Y X
= ∑∑ (xy − yμ x − xμ y + μ x μ y ) f ( x , y )
Y X
= ∑∑ xy f ( x , y ) − ∑∑ yμ x f ( x , y ) − ∑∑ xμ y f (x , y ) + ∑∑ μ x μ y f (x , y )
Y X Y X Y X Y X
= ∑∑ xy f (x , y ) − μ x ∑∑ y f ( x , y ) − μ y ∑∑ x f ( x , y ) + μ x μ y ∑∑ f ( x , y )
Y X Y X Y X Y X
= ∑∑ xy f ( x , y ) − μ x ∑ y f ( y ) − μ y ∑ x f ( x ) + μ x μ y
Y X Y X
= ∑∑ xy f ( x , y ) − μ x μ y − μ y μ x + μ x μ y
Y X
= ∑∑ xy f (x , y ) − μ x μ y
Y X
⎛ ⎞⎛ ⎞
= ∑∑ xy f ( x , y ) − ⎜ ∑ x f ( x )⎟ ⎜ ∑ y f ( y )⎟
Y X ⎝ X ⎠⎝ Y ⎠
- 32 -
In special cases where X = Y and μ x = μ y the COV ( X ,Y ) = Var ( X ) = E ( X − μ x )
2
Cov ( X ,Y )
ρ x ,y =
σ xσ y
Definition
g ( X 1 , X 2 , X 3 , X 4 .... X k ) is
then
Note
In this unit we deal with k=2. When k≥2, will be dealt with in probability & Statistic IV
Example
⎧2 x ,0 ≤ x1 ≤ 1,0 ≤ x2 ≤ 1
f ( x1 , x2 ) = ⎨ 1
⎩0, otherwise
- 33 -
Find
1. F ( x1 / x2 )
2. E ( X 1 X 2 )
3. E ( X 1 )
4. Var( X 1 )
5. Var( X 1 X 2 )
Solution
1 1
2. E ( X 1 , X 2 ) = ∫ ∫ x1 , x2 f ( x1 , x2 )d x1 d x2
0 0
1
1 1
⎡ 3⎤
= ∫ ∫ x1 , x 2 (2 x1 )d x , d x = x 2 x1 d
1
1
∫⎢ 22
⎥ x2
0 0
⎣ 3 ⎦ 0
0
1
⎛2⎞
1
2 ⎡ x2 ⎤
2
= ∫ ⎜ ⎟ x2 d x2 = ⎢
3
0⎝ ⎠ 3 ⎥
⎣ ⎦2 0
1
=
3
1 1
3. E ( X 1 ) = g (x1 , x 2 ) = x1 = ∫ ∫ x1 (2 x1 )d x1 , d x2
0 0
1
1⎡ 2 x13 ⎤ 1
2
= ∫⎢ ⎥d x2 = ∫ 3d x2
0
⎢⎣ 3 ⎥⎦ 0
0
1
2
= ∫
0
3
d x 2 =
2
3
x2 ]
1
- 34 -
2
=
3
( )
1 1
4. E X 1 = ∫∫ x1 f ( x1 , x2 )d x1 , d x2
2 2
0 0
1
1 1 ⎡ 2 x14 ⎤
1
1
⎛1⎞ ⎡ ⎤
1
= ∫ ∫ 2 x1 d x1 , d x2
3
= ∫⎢ ⎥d = ∫ ⎜ ⎟d x2 x2
2 =⎢ ⎥
0⎝ ⎠
x2
0 0
⎣ 4 ⎦
0
0
2
⎣ ⎦0
1
=
2
( )
5. Var ( X 1 ) = E x1 − (E ( x1 ))
2 2
2
1 ⎛2⎞
= −⎜ ⎟ =
1
2 ⎝3⎠ 18
6. Var ( X 1 X 2 ) = ?
Question
The random variables X1 and X2 have the joint probability distribution function given by
⎧2(1 − x1 ),0 ≤ x1 ≤ 1; 0 ≤ x2 ≤ 1
f (x1 , x2 ) = ⎨
⎩0, otherwise
Find
1. F (x1 x2 )
2. E ( X 1 X 2 )
3. E ( X 1 )
4. E ( X 2 )
5. Var( X 2 )
- 35 -
Properties of Expected Value of Random Variables
E [Cg ( X 1 , X 2 )] = CE [g ( X 1 , X 2 )]
E [g1 ( X 1 , X 2 ) + g 2 ( X 1 , X 2 ) + ⋅ ⋅ ⋅ + g k ( X 1 , X 2 )] = Eg1 ( X 1 , X 2 ) + Eg 2 ( X 1 , X 2 ) + ⋅ ⋅ ⋅ + Eg k ( X 1 , X 2 )
Definition
notation form,
Cov( X 1 X 2 ) = E[( X 1 − μ1 )( X 2 − μ 2 )]
Or
Note
The larger the value (absolute value) of the covariance of X1 and X2, the greater the linear
dependence between X1 and X2, Positive values of the covariance indicate that X1 increases as
Unfortunately it is difficult to use the covariance as a measure of dependence because, its value
depends upon the scale of measurement and therefore it is hard to determine whether a
- 36 -
This problem can be eliminated by standardizing its values, using simple coefficient of linear
correlation.
Definition
Let X1 and X2 be two random variables. The correlation coefficient between X1 and X2 is
defined as
Cov( X 1 , X 2 ) δx1 x2
ρ= =
Var( x1 ) Var(x 2 ) σxσx
1 2
Note
1. − 1 ≤ ρ ≤ 1
3. When ρ = 0 , then the covariance is zero and therefore no correlation between the two
variables.
Example
⎧3x , 0 ≤ x2 ≤ x1 ≤ 1
f ( x1 , x2 ) = ⎨ 1
⎩0, elsewhere
Find
Solution
Cov ( X 1 , X 2 ) = E ( X 1 , X 2 ) − E ( X 1 )E ( X 2 )
- 37 -
1 x1
Now E ( X 1 , X 2 ) = ∫ ∫ x1 x2 (3x1 )d x2 d x1
0 0
1 ⎡ x2 2 ⎤
= ∫ 3x1 ⎢ ⎥d
2
x1
0
⎣ 2 ⎦
1
3 ⎡ x1 ⎤ =
5 3
= ⎢ Check
2 5 ⎥ 10
⎣ ⎦ 0
3 3
E(X 1 ) = , E(X 2 ) =
4 8
3 ⎛ 3 ⎞⎛ 3 ⎞
Then Cov( X 1 , X 2 ) = − ⎜ ⎟⎜ ⎟ = 0.02 check
10 ⎝ 4 ⎠⎝ 8 ⎠
Theorem
E( X i ) = ε i
Define
n m
U1 = ∑ aiYi , and U 2 =
i =1
∑b X
j =1
i j
a) E (U 1 ) = ∑ aU i i
i =1
i =1 i j
- 38 -
∑ ∑ a b Cov (Y , X )
n m
c) Cov (U 1 , U 2 ) = i j i j
i j =1
Proof
⎛ n ⎞ n n
E (U 1 ) = E ⎜ ∑ a i Yi ⎟ = ∑ a i E (Yi ) = ∑ aU i i
⎝ i =1 ⎠ i i
Var (U 1 ) = E (U 1 ) − (E (U 1 ))
2
2
⎛ n n ⎞
= E ⎜⎜ ∑ a iY i − ∑ a iU i ⎟⎟
⎝ j =1 i=1 ⎠
2
⎛ n ⎞
= E ⎜⎜ ∑ a i (Yi − U i )⎟⎟
⎝ j =1 ⎠
⎡ n
= E ⎢ ∑ a i (Y i − U
2
i )2 + ∑∑ aia j (Y i −U i )(Y j −U j )⎤⎥
⎣ j =1 i≠ j ⎦
)(Y j )
n
∑ a i E (Y i − U )2 ∑∑ a i a j E (Y i − U
2
= i + i −U j
j =1 i≠ j
Note that
- 39 -
Cov (U 1 , U 2 ) = E [(U 1 − E (U 1 ))(U 2 − E (U 2 ))]
⎡⎛ n n
⎞⎛ m m ⎞⎤
= E ⎢ ⎜ ∑ a i Y i − ∑ a iU i ⎟ ⎜⎜ ∑ b j X j − ∑ b j ε ⎟
j ⎟⎥
⎣⎢ ⎝ i i =1 ⎠ ⎝ j =1 j =1 ⎠ ⎦⎥
⎧⎪ ⎡ ⎛ n ⎞⎛ m ⎤⎫
= E ⎨ ⎢ ⎜ ∑ a i (Yi − U i )⎟ ⎜⎜ ∑ b j (X j − ε j )⎞⎟⎟ ⎥ ⎪⎬
⎪⎩ ⎣⎢ ⎝ i ⎠ ⎝ j =1 ⎠ ⎦⎥ ⎪⎭
⎡ n m
= E ⎢ ∑ ∑ a i b j (Y i − U i )(X j − ε j )⎤⎥
⎣ i =1 j =1 ⎦
− U i )(X j − ε )
n m
= ∑ ∑ a b E (Y
i =1 j =1
i j i j
∑ ∑ a b Cov (Y , X )
n m
= i j i j
i =1 j =1
Note:
In general, multivariate normal density function would be defined for k continuous random
variables X 1 , X 2 , X 3 .... X k . For this unit, we require K=2 (bivariate) which is defined as
⎧ 1 ⎧ Q⎫
⎪ exp⎨− ⎬ , − ∞ < x1 < ∞, − ∞ < x2 < ∞
f (x1 , x2 ) = ⎨2π ,σ1σ 2 1 − ρ2 ⎩ 2 ⎭1
⎪0, elsewhere
⎩
- 40 -
In matrix form, it can be written as
⎧
⎪ 1 ⎧ 1 ⎫
exp⎨− (x −μ)1∑ (x−μ)1⎬, ……………………..…….(**)
−1
f (x) = ⎨ 1
−
⎪2π ⎩ 2 − −
⎭
⎩ ∑ 2
⎡ x1 ⎤
Where x = ⎢ ⎥,−∞ ≤ xi ≤ ∞, i = 1,2
⎣ x2 ⎦
⎡δ 11 δ 21 ⎤ ⎡μ ⎤
∑ Is a 2x2 Variance covariance matrix of X , i.e. ∑ = ⎢δ ⎥
δ 22 ⎦
and μ = ⎢ 1 ⎥
⎣ 12 ⎣μ 2 ⎦
Question
Solution
⎡δ 11 δ 21 ⎤
∑ = ⎢δ δ 22 ⎥⎦
⎣ 12
−1
1 ⎡ δ 11 − δ 12 ⎤
∑ =
δ 11δ 22 − δ 12 2
⎢− δ δ 22 ⎥⎦
⎣ 12
But
δ 12
ρ12 = ρ = ⇒ δ 12 = ρ12 δ 11δ 22
δ 11δ 22
Then
−1
1 ⎡ δ 22 − δ12 ⎤
∑ = ⎢
δ11δ 22 (1 − ρ12 2 ) ⎣− δ12 δ11 ⎥⎦
- 41 -
⎡ ( x1 − μ1 )2 (x2 − μ2 )2
1 (x1 − μ1 ) (x2 − μ2 )2 ⎤
= 2 ⎢
+ + ρ ⎥ …..(***)
1 − ρ12 ⎣⎢ σ11 σ 22
12
σ11 σ 22 ⎦⎥
∑ ( )
= δ11δ 22 − δ12 = δ11δ 22 1 − ρ122 …………………………… (3)
2
1 ⎧⎪ 1 ⎡ ( x − μ )2 ( x − μ ) 2 ⎛ ( x1 − μ 1 )2 ⎞⎛ ( x 2 − μ 2 )2 ⎞ ⎤ ⎫⎪
f ( x1 , x 2 ) = exp ⎨− ⎢ 1 1
+ 2 2
− 2 ρ 12 ⎜ ⎟⎜ ⎟⎥ ⎬
(
2π δ 11δ 22 1 − ρ 122 ) (
⎪⎩ 2 1 − ρ 12
2
) ⎢⎣ δ 11 δ 22 ⎜
⎝ δ 11 ⎟⎜
⎠⎝ δ 22 ⎟⎥
⎠ ⎦ ⎪⎭
Note:
If X1 and X2 are uncorrelated, and then the joint PDF of X1 and X2 can be written as the
Where
1 ⎡ (x1 − μ1 ) ⎛ (x1 − μ1 )( x 2 − μ 2 ) ⎞ (x 2 − μ 2 )2 ⎤
2
Q= ⎢ − 2 ρ12 ⎜⎜ ⎟⎟ + ⎥
(
1 − ρ 2 ⎣ δ12 ) ⎝ δ 1δ 2 ⎠ δ 22 ⎦
(
If ( X 1 , X 2 ) ≈ BVN μ1 , μ2 , δ1 , δ 2 , ρ
2 2
)
( )
Show that, X 1 ≈ N μ1 , δ12 , X 2 ≈ N μ 2 , δ 2 ( 2
)and ρ is the Coefficient of X and X 1 2.
Question
- 42 -
Joint Moment Generating Function
The moment generating function, discussed in probability and statistics II can be generalized to
Definition
as
⎡ ⎛ k ⎞⎤
M X (t ) = E ⎢exp⎜ ∑ t i xi ⎟⎥
⎣ ⎝ i =1 ⎠⎦
NOTE:
with respect to t i and S times with respect to t j and then setting all t i = t j = 0
- The joint MGF also uniquely determines the joint distribution of variables X 1 ,..., X k
-The MGF of marginal distributors can also be obtained from the joint MGF e.g.
M x (t 1 ) = M x, y (t 1 ,0 )
M y (t 2 ) = M x, y (0, t 2 )
M x, y (t 1 ,t 2 ) = M x (t 1 )M y (t 2 )
Question ???
Example
Suppose X and Y have density functions f ( x, y ) = λ2 e − λy if 0 < x < y < ∞ . Find the joint mgf.
Solution
- 43 -
∞ y
M x , y (t1 , t 2 ) = E (e t x +t y ) = ∫ ∫ e t x +t y λ2 e −λy dxdy
1 2 1 2
0 0
y y
∞ t1 x ∞ t1 x
e e
= λ2 ∫ e t2 y e −λy dy == λ2 ∫ e − y (λ −t2 ) dy
0
t1 0
t1
0 0
∞
e t1 y − y (λ −t2 ) 1 − y (λ −t2 )
= λ2 ∫ e − e dy
0
t1 t1
∞
λ2 − y ( λ −t1 −t 2 )
= ∫e − e − y (λ −t2 )dy
t1 0
∞
λ2 ⎡ e − y (λ −t −t ) e − y ( λ −t ) ⎤
1 2 2
= ⎢ − ⎥
t1 ⎣ − (λ − t1 − t 2 ) − (λ − t 2 ) ⎦ 0
λ2 ⎡ ⎡⎛ 1 ⎞ ⎛ 1 ⎞⎤⎤
= ⎢(0 − 0) − ⎢⎜⎜ ⎟⎟ − ⎜⎜ ⎟⎟⎥⎥
t1 ⎢⎣ ⎣⎝ − (λ − t1 − t )
2 ⎠ ⎝ − (λ − t )
2 ⎠⎦ ⎥⎦
λ2 ⎡ ⎡⎛ 1 ⎞ ⎛ 1 ⎞⎤ ⎤
= ⎢− ⎢⎜⎜ ⎟⎟ − ⎜⎜ ⎟⎥ ⎥
t1 ⎢⎣ ⎣⎝ − (λ − t1 − t 2 ) ⎠ ⎝ − (λ − t 2 ) ⎟⎠⎦ ⎥⎦
λ2 ⎛ 1 ⎞ ⎛ 1 ⎞
= ⎜⎜ ⎟−⎜ ⎟
t1 ⎝ (λ − t1 − t 2 ) ⎟⎠ ⎜⎝ (λ − t 2 ) ⎟⎠
λ2 ⎛ 1 1 ⎞
= ⎜⎜ − ⎟
t1 ⎝ (λ − t1 − t 2 ) (λ − t 2 ) ⎟⎠
Question
Find
1. k
- 44 -
3. Find the conditional density of X given y = 6
4. Find E (x y = 1) , E ( y 2 x = 2 )
Question
Find
1. k
2. E ( x )
3. E ( y )
4. Var ( x )
5. Var ( y )
7. Correlation coefficient.
Question
⎧λxe − λx , x>0 ⎫
f(x)= ⎨ ⎬
⎩0 , elsewhere⎭
Solution
∞
( )
M x (t ) = E e tx = ∫ e tx λxe −λx dx
0
∞
∞
− x ( λ −t )
⎡ xe − x (λ −t ) ∞ e − x (λ −t ) ⎤
= λ ∫ xe dx = λ ⎢ −∫ dx ⎥
0 ⎣ − (λ − t ) 0 − (λ − t ) ⎦ 0
∞
⎡ xe − x (λ −t ) e − x (λ −t ) ⎤ ⎡ ⎛ 1 ⎞⎤
= λ⎢ − 2⎥
= λ ⎢ (0 − 0 ) − ⎜ 0 − ⎟⎥
⎣ − (λ − t ) (λ − t ) ⎦ 0 ⎢⎣
⎜
⎝ (λ − t )2 ⎟⎠⎦⎥
- 45 -
⎡ ⎛ 1 ⎞⎤ λ
= λ ⎢− ⎜⎜ 0 − ⎟ =
2 ⎟⎥
⎣⎢ ⎝ (λ − t ) ⎠⎦⎥ (λ − t )2
Conditional Expectations
E [X Y ] = ∑ x f ( x y ) for x, y discrete
x
∞
=
−∞
∫ x f (x y ) dx for x, y Continuous
[
Also Var [X Y ] = E X 2 Y − (E [X Y ]) ] 2
2
⎛ ⎞
= ∑ x f (x y ) − ⎜ ∑ x f (x y )⎟
2
for x, y discrete
x ⎝ x ⎠
2
∞
⎛∞ ⎞
= ∫ x f (x y ) dx − ⎜⎜ ∫ x f (x y ) dx ⎟⎟
2
for x, y Continuous
−∞ ⎝ −∞ ⎠
Theorem 1
E [E ( X Y )] = E [X ]
⎡ ⎤
E [E ( X Y )] = ∑ ⎢∑ x f (x y )⎥ f ( y )
y ⎣ x ⎦
= ∑∑ x f (x y ) f ( y )
y x
- 46 -
f (x , y )
= ∑∑ x f ( y ) = ∑∑ x f ( x , y )
y x f (y) y x
= ∑ x∑ f (x , y ) = ∑ x f (x ) = E[X ]
x y x
Theorem 2
E [X E (Y X )] = E [ XY ]
Definition
defined as
and
∞
E [X 1 X 2 = x ] = ∫ x f (x x )dx
1 1 2 1 , if X1 and X2 are continuous
−∞
Example:
⎧1
⎪ , 0 ≤ x1 ≤ x 2 ;0 ≤ x 2 ≤ 2
f ( x1 , x 2 ) = ⎨ 2
⎪⎩0, elsewhere
Find
Solution
First find f ( x 2 )
- 47 -
x2
x2
1 1 ⎤ x2
f ( x2 ) = ∫ dx1 = x1 =
0 2 2 ⎥⎦ 0
2
⎧1
⎪ , 0 ≤ x1 ≤ x 2 ≤ 2
f (x1 x 2 ) = ⎨ x 2
⎪0,
⎩ elsewhere
Then
∞
E (x1 x 2 = 1) = ∫ x f (x1 1 x 2 )dx1
−∞
1
= ∫ x1 (1)dx1
0
Theorem
If X and Y are jointly distributed random variables and h(x, y) is a function, then
The theorem says that a joint expectation, such as the one on the left side of the equation, can
be solved by first finding the conditional expectation. E[h(x, y) x], and then finding its
Theorem
E[g(x)Y x] = g(x)E(Y x)
Example
Solution
- 48 -
By straight forward derivation we have (show this)
Note
E (Y X ) =
(n − x )P2
1 − P1
E ( X , Y ) = E (E ( XY X ))
= E [XE (Y X )]
⎡ X (n − X )P2 ⎤
= E⎢ ⎥
⎣ 1 − P1 ⎦
⎡ P ⎤
[ ( )]
= ⎢ 2 ⎥ nE ( X ) − E X 2 …………………………..(*)
⎣1 − P1 ⎦
2
( )
Now E ( X ) = nP1 and E X = Var ( X ) + (nP1 )
2
= nP1 (1 + (n − 1)P1 )
Therefore * becomes,
E ( XY ) = n(n − 1)P1 P2
Thus Cov ( X , Y ) = E ( X , Y ) − E ( X )E (Y )
= −nP1 P2
- 49 -
Example
( )
If μ1 = E ( X ) , μ 2 = E (Y ) and E Y x is a linear function of x ,
Show that
δ2
E (Y x ) = μ 2 + ρ (x − μ1 ) and E X (Var (Y x )) = δ 2 2 (1 − ρ 2 )
δ1
Solution
Suppose E (Y x ) = ax + b then
μ 2 = E (Y ) = E X (E (Y x )) = E X (ax + b ) = aμ1 + b
Now
Cov ( X , Y ) δ XY δ
a= = 2 = ρ 2 Where δ XY = ρδ 1δ 2 and
Var ( X 1 ) δ1 δ1
δ2
b = E (Y ) − aE ( X ) = μ 2 − ρ μ1
δ1
Then
E (Y x ) = ax + b
δ2 δ
=ρ x + μ 2 − ρ 2 μ1
δ1 δ1
δ2
= μ2 + ρ ( x − μ1 )
δ1
δ XY = E[( X − μ1 )(Y − μ2 )]
= E [( X − μ1 )(Y )] − 0
= E X {E [( X − μ1 )(Y X )]}
= E X [( X − μ1 )E (Y X )] = E X [( X − μ1 )(ax + b )]
- 50 -
= aδ 12
{(
E X [Var (X Y )] = E X E Y 2 X − (E (Y X )) ) 2
}
{( )
= E Y 2 − E X (E (Y X ))
2
}
( ) {
= E Y 2 − (E (Y )) − E X (E (Y X )) − (E (Y ))
2 2 2
}
= Var (Y ) − VarX (E (Y X ))
δ2
= Var (Y ) − VarX [ μ2 + ρ ( X − μ1 )]
δ1
δ 22 2
= Var (Y ) − ρ δ1 2
δ12
(
= δ2 1− ρ 2
2
)
Theorem
( )
Let X, Y be random Variables with E ( X ) = μ1 and E (Y ) = μ 2 if E Y X is a linear function
of x , Show that
E (Y x ) = μ 2 + ρ
δ2
δ1
(
(x − μ1 ) and E X (Var (Y x )) = δ 2 2 1 − ρ 2 )
Proof
( )
If E Y x = ax + b , then
δ XY = E[( X − μ1 )(Y − μ 2 )]
= E [( X − μ1 )(Y )] − 0
= E X {E [( X − μ1 )(Y X )]}
- 51 -
= E X [( X − μ1 )E (Y X )] = E X [( X − μ1 )(ax + b )]
= aδ 12
Thus
Cov ( X , Y ) δ XY δ
a= = 2 = ρ 2 and
Var ( X 1 ) δ1 δ1
δ2
b = E (Y ) − aE ( X ) = μ 2 − ρ μ
δ1 1
i.e. E (Y x ) = ax + b
δ2 δ
=ρ x + μ 2 − ρ 2 μ1
δ1 δ1
δ2
= μ2 + ρ ( x − μ1 )
δ1
NOTE:
Theorem:
(
If ( X 1 , X 2 ) ≈ BVN μ1 , μ 2 , δ 1 , δ 2 , ρ
2 2
)
Then
i) conditional on X 1 = x1 ,
⎛ δ
(
X 2 X 1 = x1 ≈ N ⎜⎜ μ 2 + ρ 2 ( X 1 − μ1 ), δ 2 1 − ρ 2
δ1
2
)⎞⎟⎟
⎝ ⎠
- 52 -
ii) Conditional on X 2 = x2 ,we have
⎛ δ
(
X 1 X 2 = x 2 ≈ N ⎜⎜ μ1 + ρ 1 ( X 2 − μ 2 ), δ 1 1 − ρ 2
δ2
2
)⎞⎟⎟
⎝ ⎠
Show this!!!
Note
δ1
E ( X 1 X 2 ) = μ1 + ρ ( X 2 − μ 2 ) is sometimes referred to as regression function of
δ2
X 1 X 2 Multiple?
Example
3
μ1 = μ 2 = 2, δ1 = δ 2 = 2 and ρ = ,
5
Calculate
i. ρ ( X 1 > 4)
(
ii. ρ X 1 > 4 X 2 = 3 )
Solution
⎛ X1 − 2 4 − 2 ⎞
ρ ( X 1 > 4) = P⎜ > ⎟
⎝ 2 2 ⎠
= P(Z > 1) = ?
δ1
ii. μ = μ1 + ρ (X 2 − μ2 )
δ2
3⎛ 2⎞
= 2 + ⎜ ⎟(3 − 2 ) = ?
5⎝2⎠
- 53 -
and
δ = δ1 1 − ρ 2
9
= 2 1−
25
= 1.6
⎛ 1.4 ⎞
= P⎜ Z > X 2 = 3⎟ = ?
⎝ 1.6 ⎠
Suppose we are given the joint PDF of X 1 , X 2 … X p and we wish to determine the joint
Y p = g p (X 1 , X 2 … X p )
Then we find the joint distribution of Y1 , Y2 ⋅ ⋅ ⋅ Y p and finally find the marginal distribution of
following transformation
Y1 = g 1 (X 1 , X 2 … X p ),Y2 = g 2 (X 1 , X 2 … X p )⋅ ⋅ ⋅ Y p = g p (X 1 , X 2 … X p ) ,
- 54 -
The Jacobian of the transformation becomes
⎡ δX 1 δX 1 δX 1 ⎤
⎢ ..... ⎥
⎢ δY1 δY2 δY p ⎥
⎢ . . . ⎥
J =⎢
⎢ . . . ⎥⎥
⎢ δX p δX p δX p ⎥
⎢ δY ......
⎣ 1 δY2 δY p ⎥⎦
⎧⎪ f (w1 (Y ), w2 (Y ),...w p (Y ) ) J , Y ∈ R p
g (Y1 , Y2 ,......, Y p ) = ⎨
⎪⎩0, elsewhere
⎡ y1 ⎤
⎢ ⎥
⎢ y2 ⎥
Where y = ⎢. ⎥
⎢ ⎥
⎢. ⎥
⎢y ⎥
⎣ p⎦
Note
Example
⎧2(1 − x1 ), 0 ≤ x1 ≤ 1; 0 ≤ x 2 ≤ 1
f ( x1 , x 2 ) = ⎨
⎩0, elsewhere
- 55 -
Solution
dX 1 dX 1
J = dv du
dX 2 dX 2
dv du
1 0 1
= 1 1 =
− 2 V
V V
⎧ f (w1 (V ,U ), w2 (V ,U )) J ; V ,U ∈ R
f (V ,U ) = ⎨
⎩0, elsewhere
⎧ 1
⎪2(1 − V ) , 0 ≤ V ≤ 1; 0 ≤ U ≤ V , or 0 ≤ U ≤ V ≤ 1
f (V , U ) = ⎨ V
⎪⎩0, elsewhere
The PDF of U is
α
fU (U ) = ∫α f (v, u )dv
−
1 1
1 ⎛1 ⎞
= ∫ 2(1 − v ) dv == 2 ∫ ⎜ − 1⎟dv
u⎝
u
v v ⎠
= 2⎛⎜
⎝ [ln v −v] ⎞⎟⎠ = 2{ln 1 − 1 − (ln u − u )}
1
⎧2{u − ln u − 1}, 0 ≤ u ≤ 1
=⎨
⎩0, elsewhere
- 56 -
α
E (U ) = ∫ uf u (u )du
−α
α
= ∫ 2u (u − ln u − 1)du
−α
⎧1 1 1
⎫
= 2⎨∫ u 2 du − ∫ u (ln u )du − ∫ udu ⎬
⎩0 0 0 ⎭
⎧⎪ u 3 ⎤ 1 1 u 2 ⎤ ⎫⎪
1
= 2⎨ ⎥ − ∫ u (ln u )du − ⎥ ⎬
⎪⎩ 3 ⎦ 0 0 2 ⎦0 ⎪
⎭
1
1
⎡u 2 ⎤ 1 u2 1
∫0 u(ln u )du = ⎢⎣ 2 (ln u )⎥⎦ − ∫0 2 . u du
0
1
⎡ u2 ⎤ 1
= ⎢0 − ⎥ = −
⎣ 4 ⎦0 4
Thus
⎡1 ⎛ 1 ⎞ 1 ⎤ ⎡1⎤ 1
E (U ) = 2⎢ − ⎜ − ⎟ − ⎥ = 2⎢ ⎥ =
⎣3 ⎝ 4 ⎠ 2 ⎦ ⎣12 ⎦ 6
Question
⎧2 x , 0 ≤ x1 ≤ 1; 0 ≤ x 2 ≤ 1
f (x1 , x 2 ) = ⎨ 1
⎩0, elsewhere
- 57 -
Question
⎧e − ( x1 + x2 ) ,0 ≤ x1 ,0 ≤ x 2
f ( x1 , x 2 ) = ⎨
⎩0 elsewhere
T –Distribution
If X ≈ N μ , σ(2
)
X −μ
Then Z =
σ
If X , is the mean of the sample X 1 , X 2 ,⋅ ⋅ ⋅ X n drawn randomly for normal population with mean
X −μ
Z=
σ
n
S =∑
(X − X ) 2
X −μ
i.e. Z = S
n −1
n
If n is small and σ is unknown then we have a t- distributed random variable with n-1 degrees
X −μ
of freedom, i.e. t = S
n
Note
- 58 -
Then to test the hypothesis
H 0 : μ = μ 0 → (specificValue )
1.
H 1 : μ > μ 0 → (Upper − tail )
Or
H 0 : μ = μ 0 → (specificValue )
2.
H 1 : μ < μ 0 → (lower − tail )
Or
H 0 : μ = μ 0 → (specific Value )
3.
H 1 : μ ≠ μ 0 → (Use − 2 − tailed )
We calculate
X −μ
t= and
S
n
Suppose that independent random samples are selected from each of two normal populations ;
X 11 , X 12 , X 13 ... X 1n1 ,from the first and X 21 , X 22 , X 23 ... X 2 n2 from the second ,where the
mean and variance of the i th population are μi and δ i 2 ,i=1,2. Further assume that Xi
2
and S i , i=1, 2 are the corresponding sample means and variances.
- 59 -
1 n1
1 n2
X1 =
n1
∑ X 1i and X 2 =
i n2
∑X
i
2i
The unbiased estimation of the variance is obtained by pooling the sample data to obtain
∑ (X ) ( )
n1 n2
− X 1 + ∑ X 2i − X 1
2 2
1i
S2 = i i
n1 + n2 − 2
=
(n1 − 1)S1 + (n2 − 1)S 2
2 2
n1 + n2 − 2
t=
(X 1 )
− X 2 − (μ1 − μ 2 )
1 1
S +
n1 n2
To test the null hypothesis H 0 : μ1 − μ 2 = D0 for some fixed value Do, it follows that if Ho is
t=
(X 1 − X 2 − (D0 ) )
1 1
S +
n1 n2
Example
Standard procedure
n1 = 9 n2 = 9
X 1 = 35.22 sec onds X 2 = 31.56 sec onds
∑ (X ) ∑ (X )
9 9
2 2
1i −X = 195.56 2i − X2 = 160.22
i =1 i =1
- 60 -
Test the hypothesis that the two populations have the same mean.
Solution
H 0 : μ 1 − μ 2 = 0 , against
H 1 : μ1 − μ 2 ≠ 0
t=
(X 1 )
− X 2 − (D0 ) D
, 0 =0
1 1
S +
n1 n2
Now
n1 n2
∑ ( X 1i − X 1 ) 2 + ∑ ( X 2i − X 1 ) 2
S2 = i i
n1 + n2 − 2
S2 =
(n1 − 1)S12 + (n2 − 1)S 2 2
n1 + n2 − 2
=
(195.56) + (160.22 ) = 22.24
9+9−2
⇒t=
(X− X2
1
=
)
35.22 − 31.56
= 1.65
1 1 ⎛ 1 1⎞
S + 4.71 ⎜ + ⎟
n1 n2 ⎝9 9⎠
The tabulated t value is t 0 .025 ,16 = 2 . 120 since t calculated =1.65< t 0 .025 ,16 = 2 . 120 we do not
reject Ho this implies there is not sufficient to indicate a difference in the two procedures.
Question
The strength of concrete depends, to some extent, on the method used for drying. Two different
drying methods showed the following results for independently tested specimens
- 61 -
Method 1 Method 2
N1 = 7 N 2 = 10
X 1 = 3250 X 2 = 3240
S 1 = 210 S 2 = 190
Use α = 0.05 .
Paired T-Test
Example
An industry in deciding whether to purchase a machine of design A or B, checks the time for
completing a certain task on each machine. Nine technicians were used in the experiment, with
each technician using both machines A and machine B in a randomized order. The time (in
Technicians 1 2 3 4 5 6 7 8 9
Test if there is a significant difference between the completion times at the 5% significance
level.
Solution
- 62 -
Sample 1 2 3 4 5 6 7 8 9
−
d − μd
td =
sd / n
Where d =
−
∑ di
=
0 .5
= 0 . 056
n 9
n
(
⎛ di − d 2 ) ⎞⎟ = 0 . 002778
sd
2
= ∑ ⎜⎜ ⎟
i =1 ⎝ n −1 ⎠
Then
0.056
td = = 3.17
0.053 / 9
t0.025 ,8 = 2.306
Conclusion:
Question:
- 63 -
Question
Consider an experiment to test the effects of a particular drug on human pulse rate. Six subjects
are chosen and their pulse rates measured both before and after the treatment, with the
following results.
Subjects 1 2 3 4 5 6
Before 73 69 70 64 69 66
After 78 73 70 69 68 72
Do the pulse rates taken after the stimulus differ significantly from those taken before it?
Take α = 0.05 .
Solution
two pulses.
xd 5 4 0 5 -1 6
1 6 (5 + 4 + ... + 6) = 3.17
xd = ∑
6 1
xi =
6
2
sd =
1 6
∑
5 1 5
[ ]
(xi − x )2 = 1 (5 − 3.17 )2 + ⋅ ⋅ ⋅ + (6 − 3.17 )2 = 2.932
s d = 2.93
The tabulated value, t5 ,0.025 = 2.571 since t=2.64> t5 ,0.025 = 2.571 reject H 0 .
Conclusion:
- 64 -
The effect of the treatment on pulse rates is significant. It is reasonable to conclude that there
Assume that we have a random sample X 1 , X 2 , X 3 .... X n from a normal distribution with
Then under H 0 ,
X2 =
(n − 1)S 2 has a χ 2 distribution with n − 1 degree of freedom.
δ 02
H 0 : σ 2 = δ 0 against
2
i.
H1 : σ 2 > δ 0 (upper − tail )
2
Or
H 0 : δ 2 = δ 0 against
2
ii.
H1 : δ 2 < δ 0 (lower − tail )
2
Or
H 0 : δ 2 = δ 0 against
2
iii.
H1 : δ 2 ≠ δ 0
2
X2 =
(n − 1)S 2
δ 02
- 65 -
For (i) reject H 0 if X calculated >
2
χα2 , n −1
Example
A machine engine part produced by a company is claimed to have diameter variance no larger
Solution
Assume the measured diameters are normally distributed. The test statistic
(n − 1) s 2
X2 =
δ 02
9(0.0003)
= = 13.5
0.002
Conclusion:
Question
An experimenter was convinced that his measuring equipment possessed variability, which
the data disagree with his claim? Take a = 0.01 . What would you conclude if you
choose a = 0.05 .
- 66 -
F-Test
We may be interested in comparing the variance of two normal distributions, and testing to
determine whether or not they are equal. This problem is often encountered when comparing
manufactured product, or the variation in scores for two testing procedures. Example, suppose
x11 , x12 , x13 ....x1n1 and x21 , x22 , x23 ....x2 n2 are independent random samples from normal
distributions with unknown means and var( X 1i ) = σ 12 , var( X 2 i ) = σ 2 2 , where σ 1 and σ 2 are
2 2
unknown. Suppose we are interested to test the null hypothesis, H 0 : σ 1 = σ 2 against the
2 2
H1 : σ 1 > σ 2
2 2
now, σ 1 and σ 2 can be estimated by S1
2 2 2
alternative hypothesis and
2
S 2 respectively
2 2
We would reject H 0 in favour of H1 if S1 is much larger than S 2 i.e. reject H 0 if
(n1 − 1 )s1 2
δ 1 2 (n1 − 1 ) ……………..***
F =
(n 2 − 1 )s 2 2
δ 2 2 (n 2 − 1 )
s1 δ 2
2 2
= >k
s2 δ1
2 2
2
s1
= 2
s2
- 67 -
2
s
Denominator degrees of freedom. Under the null hypotheses δ1 = δ 2
2 2
, then F = 1 2 has an
s2
Example
Consider two random samples, X1 and X2 of sizes 10 and 20 with sample variances given as
0.0003 and 0.0001 respectively. Assuming that the populations, from which the samples have
been drawn, are normal, determine whether the variance of the first population is significantly
Solution
Let δ 12 and δ 2 denote the variances for the first and second population from which the
2
2
s1
F= 2 , based on V1=9 and V2=19 d.f
s2
Now
2
s1 0.0003
F= 2
= =3
s2 0.0001
Since F-calculated > F9,19, 0.05 = 2.42 we reject the null hypothesis.
Conclusion
The variation of the first population is greater than the second one.
- 68 -
Note:
1
X = ∑ xi is normally distributed with a mean of µ and a
n i
δ2
variance of .
n
Show that……………………………
Xi − μ
2. Suppose X1, X2, X3…Xn is as defined in 1. then Z i = are independent standard
δ
2
⎛ Xi − μ ⎞
n n
normal random variables i=1,2, ……n and ∑ Z i = ∑ ⎜
2
⎟ , is a chi-square
i i ⎝ δ ⎠
3. Let Z be a standard normal random variable and let χ v2 be a chi-square random variable
4. Let χ 1 and χ 2 be chi-square random variables with V1 and V2 d.f, respectively. Then
2 2
χ12 / V1
F= 2 , is said to have an f distribution with V1 numerator
χ 2 / V2
- 69 -
Question
Consider two random samples X1 and X2 of sizes 9 and 5 with sample variances 115 and 24
respectively.
Assuming that the populations, from which the samples have been drawn, are normal,
determine whether the samples could have come from a population with a common variance.
Question
Eight students took two complete sciences practical in successive weeks, and obtained the
1 12 11
2 12 11
3 13 15
4 10 11
5 12 12
6 14 10
7 13 14
8 10 12
Assuming that the marks are normally distributed, carry out a paired sample t-test to determine
whether there is a significant difference between performance in the first and second practical.
Question
- 70 -
⎡ 25 − 12⎤
∑ = ⎢− 12 16 ⎥⎦
⎣
Determine the standard deviation of X and of Y. Also the correlation coefficient between x and
Ordered Statistics
x(1) ≤ x( 2) ≤ x(3) ≤ .... ≤ x( n ) (for continuous random variables, equality signs can be
ignored)
i.e.
x ( 1 ) = Min (x 1 , x 2 , x 3 .... x n )
is the minimum of Xi’s and x ( n ) = Max (x 1 , x 2 , x 3 .... x n ) the maximum of the Xi’s
Question
Solution
i = 1 , 2 , 3 .... n i.e.
That
- 71 -
P(X (n) ≤ x) = P ( X 1 ≤ x)P ( X 2 ≤ x)P ( X 3 ≤ x )...... P ( X n ≤ x)
= [F (x)] n
sides, we get
f n ( x) = n[F ( x ) ]
n −1
f ( x)
Question
Solution
(X (1 ) > x) occur iff the events ( X i > x ) occur for i = 1 , 2 , 3 .... n .Because
= 1 − [1 − F ( x )]
n
Let f1 ( x ) denote the density of X (1) , then differentiating both sides, we get
f1 ( x ) = n[1 − f ( x ) ]
n−1
f ( x)
- 72 -
Example:
A computer component has length of life X, measured in hours, with probability function
⎧⎛ 1 ⎞ 100 −x
Suppose that two such components operate in parallel, i.e. the computer does not fail until both
components fail.
Find
2. the median of Y
Solution
−x
1 100
But f ( x) = e
100
x −t
1
⇒ F ( x) = ∫0 100 l 100 dt
x
−t
⎤
= −l 100
⎥
⎦0
−x
= 1− l 100
f Y ( x ) = n[F ( x )]
n −1
f ( x)
⎡ −x
⎤ 1 100−x
= 2 ⎢1 − e ⎥
100
e ,x > 0
⎣ ⎦ 100
- 73 -
⎧⎛ 1 ⎞⎛ − x −x
⎞
⎪⎜ ⎟⎜⎜ e − e 50 ⎟⎟, x > 0
100
= ⎨⎝ 50 ⎠⎝ ⎠
⎪
⎩0.........................elsewhere
1 ⎛⎜ −100 ⎞
x t t
−
FY ( x ) = ∫ ⎜ e − e 50 ⎟
⎟ dt
0
50 ⎝ ⎠
1 ⎧⎪ ⎛⎜ −100 ⎞ ⎫⎪
x t x t
−
= ⎨∫ ⎜ e dt − ∫ e 50
dt ⎟⎬
⎟⎪
50 ⎪⎩ 0 ⎝ 0 ⎠⎭
⎧ ⎤
x
⎡ − ⎤ ⎪
x
⎫
1 ⎪
t t
−
= ⎨− 100 e 100
⎥ − ⎢ − 50 e ⎥ ⎬
50
50 ⎪ ⎦ 0 ⎢⎣ ⎦ 0 ⎪⎭
⎩
1 ⎪
⎧
−
x ⎡ −
x ⎤ ⎫⎪
= ⎨ ( −100 e 100 − ( −100 e ) − ⎢ − 50 e 50 − ( −50 e ) ⎥ ⎬
0 0
50 ⎪ ⎢ ⎥⎪
⎩ ⎣ ⎦⎭
1 ⎧⎪ −
x
−
x ⎫⎪
= ⎨100 −100 e 100 + 50 e − 50 ⎬
50
50 ⎪⎩ ⎪⎭
1 ⎧⎪ −
x
−
x ⎫
⎪
= ⎨ 50 + 50 e −100 e
50 100 ⎬
50 ⎪⎩ ⎪⎭
x x
− −
= 1+ e 50
− 2e 100
x 0.5 x 0.5
− −
0 .5 = 1 + e 50
− 2e 100
x0.5
−
Let m = e 100
Then 0.5 = 1 + m 2 − 2m
- 74 -
= m 2 − 2m + 0.5 = 0
b 2 − 4ac
m = −b ±
2a
2 2 − 4 (1 )(0 . 5 )
= +2 ±
2
2
= +2 ± = 1 . 707 or 0 . 293
2
x 0 .5
−
Now m = e 100
= 0.293
x0.5 = 122.758
Question
A computer component has length of life X, measured in hours, with probability density
function,
⎧⎛ 1 ⎞ −100 x
Suppose that two such components operate independently and in series a certain system i.e.
Find
2. the median of Y
Solution
Because the system fails at the first component failure Y = Min ( X 1 , X 2 ) Where X 1 and
f Y ( x ) = n[1 − F ( x )]
n −1
f ( x)
- 75 -
n −1
⎡ −
x
⎤ 1 −100
x
= 2 ⎢1 − (1 − e 100 ) ⎥ e
⎣ ⎦ 100
⎧⎛ 1 − x ⎞
⎪⎜ e 50 ⎟⎟, x > 0
= ⎨⎜⎝ 50 ⎠
⎪
⎩0............., elsewhere
The median of Y,
E Y2 (var (Y1 Y2 )) = σ 12 1 − ρ 2 . ( )
Solution
{(
EY2 (var (Y1 Y2 )) = EY2 E Y12 Y2 − (E (Y1 Y2 )) ) 2
}
{( )
= E Y12 − EY2 [E (Y1 Y2 )]
2
}
( ) {
= E Y12 − (E (Y1 )) − EY2 [E (Y1 Y2 )] − EY2 (E (Y1 Y2 ))
2 2
}
= Var (Y1 ) − VarY2 [E (Y1 Y2 )]
⎧ δ ⎫
= Var (Y1 ) − VarY2 ⎨μ1 + ρ 1 (Y2 − μ 2 )⎬
⎩ δ2 ⎭
δ 12δ 22
= Var (Y1 ) − ρ 2
δ2
(
= σ 12 1 − ρ 2 )
- 76 -
2. Suppose Y1 and Y2 have bivariate normal distribution with parameters
Solution
δ1
μ = μ1 + ρ (Y2 − μ 2 )
δ2
3 4
= 2+ ⋅ (3 − 2)
4 4
= 2.75
σ 2 = σ 12 (1 − ρ 2 )
⎛ 9⎞
= 16⎜1 − ⎟
⎝ 16 ⎠
=7
⎛Y − μ 4− μ ⎞
P(Y1 > 4 Y2 = 3) = P⎜ 1 > ⎟
⎝ σ σ ⎠
⎛ 4 − 2.75 ⎞
= P⎜⎜ Z > ⎟⎟
⎝ 7 ⎠
= P (Z > 0.472 )
= 0.3192
- 77 -
⎧ λ y 2 e −2 λ
⎪ , y1 = 0,1,2... y 2
⎪⎪ y1 ! ( y 2 − y 1 )!
p (Y1 , Y2 ) = ⎨ y 2 = 0,1,2,....
⎪0, Otherwise
⎪
⎪⎩
Solution
P (Y1 = y1 , Y2 = y 2 )
P (Y1 Y2 = y 2 ) =
P (Y2 = y 2 )
P (Y2 = y 2 ) = ∑ P( y , y ) 1 2
all Y1
y2
λ y e −2 λ
2
= ∑ y !( y
y1 = 0 − y1 )!
1 2
λ y e −2 λ
2 y2
y2!
= ∑ y !( y
y2! y1 = 0 1 2 − y 1 )!
y2
⎛ y2 ⎞
but ∑ ⎜⎜ y ⎟⎟ = (1 + 1)y 2 = 2 y 2
y1 = 0⎝ 1 ⎠
P(y2 ) =
(2λ )
y
e −2 λ
2
, y 2 = 0 ,1, 2 ...
thus
y2!
λ y e −2 λ
2
P (Y1 , Y2 ) y1!( y 2 − y1 )!
P (Y1 Y2 ) = =
P (Y2 ) (2λ ) y2 e −2 λ
y2!
y2
⎛1⎞ y2!
=⎜ ⎟
⎝2⎠ y1!( y 2 − y1 )!
- 78 -
y2 − y1
⎛ y2 ⎞ ⎛ y2 ⎞⎛ 1 ⎞ 2 ⎛ 1 ⎞
y2 y
⎛1⎞
=⎜ ⎟ ⎜⎜ ⎟⎟ = ⎜⎜ ⎟⎟⎜ ⎟ ⎜ ⎟
⎝2⎠ ⎝ y1 ⎠ ⎝ y1 ⎠⎝ 2 ⎠ ⎝ 2 ⎠
1
This is a binomial probability distribution with parameters Y2 and .
2
4. Suppose Y1 and Y2 are independent and exponentially distributed random variables with
2 parameters.
⎧⎪ 1 e − 2 ( y1 + y2 ) y1≥0 , y2 ≥0
1
( )
f y1 , y 2 = ⎨ 4
⎪⎩0 elsewhere
Solution
1
Y1 = (U + W )
2
1
Y2 = (W − U ) = − 1 (U − W )
2 2
∂Y1 ∂Y1 1 1
J = ∂U ∂W = 2 2 =1
∂Y2 ∂Y2 1 1 2
−
∂U ∂W 2 2
f (U , W ) = f (w1 (U , W ), w2 (U , W )) ⋅ J
⎧ 1 − 12 w ⎛ 1 ⎞
⎪ e ⎜ ⎟
= ⎨4 ⎝2⎠
⎪0,
⎩ otherwise
- 79 -
⎧ 1 − 12 w
⎪ e , 0 ≤W ≤U
= ⎨8 .
⎪0,
⎩ otherwise
⎧e− ( x1 + x2 ) , 0 ≤ x1; 0 ≤ x2
f ( x1 , x2 ) = ⎨
⎩0, elsewhere
Solution
Let Y2 = X 2
Therefore X 1 = Y1 − X 2 = Y1 − Y2
∂X 1 ∂X 1
∂Y1 ∂Y2 1 −1
J= = =1
∂X 2 ∂X 2 0 1
∂Y1 ∂Y2
Limits 0 ≤ x1 ; 0 ≤ x2
0 ≤ Y1 − Y2 ; 0 ≤ Y2
Y2 ≤ Y1 Hence 0 ≤ Y2 ≤ Y1
- 80 -
Y1
Y1
⎧Y1e −Y1 ,0 ≤ Y1
g (Y1 ) = ⎨
⎩0 , otherwise
⎧2 x , 0 ≤ x1 ≤ 1; 0 ≤ x2 ≤ 1
f ( x1 , x2 ) = ⎨ 1
⎩0, elsewhere
Solution
Let Y2 = X 2
Y1 Y
Therefore X 1 = = 1 and X 2 = Y2
X 2 Y2
∂X 1 ∂X 1
1 Y1
∂Y1 ∂Y 2 1
J= = Y2 − Y 22 =
∂X 2 ∂X 2 Y2
0 1
∂Y1 ∂Y 2
⎧ Y1 1
⎪2 ⋅
= ⎨ Y2 Y2
⎪0 elsewhere
⎩
Limits
0 ≤ x1 ≤ 1; 0 ≤ x2 ≤ 1
- 81 -
Y1
0≤ ≤ 1; 0 ≤ Y2 ≤ 1
Y2
Hence 0 ≤ Y1 ≤ Y2 ≤ 1
Y1
1 1 1
Y1 1
= ∫2 2
⋅dY2 = 2Y1∫ 2
⋅dY2 = 2Y1 ∫ Y2 dY2
−2
Y1 Y2 Y1 Y2 Y1
1 1
⎡ Y −1 ⎤ ⎡1⎤
= 2Y1 ⎢ 2 ⎥ = −2Y1 ⎢ ⎥
⎣ − 1 ⎦ Y1 ⎣ Y2 ⎦ Y1
1
⎡1⎤ ⎧ 1⎫
= −2Y1 ⎢ ⎥ = −2Y1 ⎨1 − ⎬
⎣ Y2 ⎦ Y1 ⎩ Y1 ⎭
⎧2(1 − Y1 ) ,0 ≤ Y1
g (Y1 ) = ⎨
⎩0 , otherwise
1
E (Y1 ) = ∫ Y1 g (Y1 )dY1
0
1
= ∫ Y1.2(1 − Y1 )dY1
0
1
⎡ Y3⎤ 4
= 2 ⎢Y12 − 1 ⎥ =
⎣ 3 ⎦0 3
Question
- 82 -
2) Find P(0 < x < 13 , 1
2 < y < 1)
4) Find E ( x ) , E ( y ) , E (x 2 ) and E ( y 2 )
Question
1) Find k
Question
Find the probability density function of Y = X 1 + X 2 if the joint probability distribution density
of X 1 and X 2 is
- 83 -
Examination for April 2006
Question 1
i) a). The joint probability distribution of two random variables X1 and X2 is shown in the
following table
Find
b). A soft drink machine has a random amount Y2 in supply at the beginning of a given day and
dispenses a random amount Y1 during the day (with measurement in gallons). It is not re-
supplied during the day and hence and hence Y1< Y2 have joint density
⎧ 12 , 0 ≤ y1 ≤ y2 , 0 ≤ y2 ≤ 2
f ( y1 , y2 ) = ⎨
⎩0, elsewhere
Find
1
ii) The probability that less than 2
gallon is sold, given that the machine contains 1 gallon
iii) P (Y1 ≥ 12 Y2 ≤ 1
4 ) 8 mks
iv) P (Y1 ≤ 12 Y2 = 2 )
c). Suppose that the random variable x1 and x2 have the joint probability density function
- 84 -
⎧12 x x (1 − x1 ), 0 ≤ x1 ≤ 1, 0 ≤ x2 ≤ 1
f ( x1 , x2 ) = ⎨ 1 2
⎩0, elsewhere
d). The random variables X and Y have chi-square distributions with n and m degrees of
freedom respectively where n > m. Find the distribution of X-Y using the method of moment
generating function.
e). If two random variables are independent, are they also un-correlated? Is the converse true?
i) Bivariate distribution
b). X1 and X2 are normally distributed random variables with means and standard deviations
μ1 , δ1 and μ 2 , δ 2 respectively.
i) Using the method of moment generating function, find the probability distribution
3. a). X1and X2 are independent random variables each with a chi-square distribution with r=2
(Recall that if X is a chi-square random variable with r degrees of freedom then its probability
( 12 ) r
2
x
r
2 −1
e
−x
2 ∞
where Γ ( x ) = ∫ xα −1e − x dx 10mks
Γ (r 2 ) 0
- 85 -
b). Let X1, X2 be a random sample from a distribution having the probability density function
⎧e x , 0 < x < ∞
f (x ) = ⎨
⎩0, elsewhere
−
1
102
[
(x + 2)2 − 2.8(x + 2)( y − 1) + 4( y − 1)2 ]
Find the values of
b). in a certain population of married couples the height X of the husband and the height Y of
the wife have a bivariate normal distribution with parameters μ1 = 5.8 units and μ 2 = 5.3
units. The standard deviations δ1 = δ 2 = 0.2 , and correlation coefficient ρ = 0.6 . Find the
probability that the height of the wife lies between 5.28 and 5.92 units given that the height of
5. a). Let U and V be two independent chi-square random variables with respective means r1
and r2.
U r1
F=
V r2
U r1
iv. Introducing Z=V and F = find the joint density of f and z.
V r2
- 86 -
v. Write down an expression for finding the marginal density of F. 10mks
b). let X be a standard normal random variable and Y a chi square random variable. X and Y
are independent.
X
i. State the distribution of
Y k
X
ii. Consider T = . Introduce Z=Y and find the joint distribution of T and Z.
Y k
iii. Write down an expression for finding the marginal density of T. 10 mks
CAT QUESTIONS
Qz
2. If two random variables X and Y have the joint probability distribution function
a) Show that P( x, y ) satisfy the properties of a discrete joint distribution function 4mks
d) Find F (2,1)
e) Find:
- 87 -
ii. The marginal distribution of Y
Qz
If X is the proportion of persons who will respond to one kind of mail-order solicitation, Y is
the proportion of persons who will respond to another kind of mail-order solicitation and the
Find
Qz
Check for each of the following probability densities whether the two random variables are
independent:
- 88 -
CAT III
SMA 2231
Q1. Let x1 , x 2 − − x n be a random sample from a normal population with mean μ and
1 n
variance σ 2 . Show that the sample mean x= ∑ xi and
n i
the sample variance
n
( x i − x )2
S2 = ∑ are independent (use sample size n=2).
i n −1
n
variance σ 2 . Let Q = ∑ ( xi − x ) where x is the sample mean. Find E (Q ) .
2
Q3. The following are observations 9random) from a normal population with mean 22 and
variance 10. They are 25, 17, 23, 20, 18, 15, 24, and 21. Calculate a statistic that is a function
Q4.
1
random sample from the distribution f ( x ) =
−x
e θ,x > 0.
θ
- 89 -