3.

5 Joint moment generating functions

The moment generating function concept can be generalized to two or more random variables.

3.5.1 Defn
The joint moment generating function of n random variables 𝑋1 , 𝑋2 , … , 𝑋𝑛 is defined as
𝑀𝑋1 ,𝑋2 ,…,𝑋𝑛 (𝑡1 , 𝑡2 , … , 𝑡𝑛 ) = 𝐸(𝑒 𝑡1 𝑋1 +𝑡2𝑋2 +⋯+𝑡𝑛𝑋𝑛 )
where −ℎ < 𝑡𝑖 < ℎ , 𝑖 = 1, 2, … , 𝑛 for some ℎ > 0.
Note:
1. Marginal moment generating functions can be obtained from the joint m.g.f. by letting all
but one of the 𝑡𝑖 ′𝑠 be 0 i.e.
𝑀𝑋𝑖 (𝑡) = 𝐸(𝑒 𝑡𝑋𝑖 ) = 𝑀𝑋1 ,𝑋2 ,…,𝑋𝑛 (0, 0, … , 0, 𝑡, 0, … , 0)
where 𝑡 is in the 𝑖 𝑡ℎ place e.g. if 𝑀𝑋,𝑌 (𝑡1 , 𝑡2 ) = 𝐸(𝑒 𝑡1 𝑋+𝑡2 𝑌 ) then
𝑀𝑋 (𝑡1 ) = 𝑀𝑋 ,𝑌 (𝑡1 , 0)
𝑀𝑌 (𝑡2 ) = 𝑀𝑋 ,𝑌 (0, 𝑡2 )
or simply
𝑀𝑋 (𝑡) = 𝑀𝑋 ,𝑌 (𝑡, 0)
𝑀𝑌 (𝑡) = 𝑀𝑋 ,𝑌 (0, 𝑡)
.
2. The joint mgf has properties similar to those of the univariate mgf e.g. for 𝑀𝑋,𝑌 (𝑡1 , 𝑡2 )
𝜕 𝑟+𝑠
𝐸(𝑋 𝑟 𝑌 𝑠 ) = 𝑟 𝑠 𝑀𝑋,𝑌 (𝑡1 , 𝑡2 ) |
𝜕𝑡1 𝜕𝑡2 (𝑡1 , 𝑡2 ) = (0,0)
2
𝜕
𝑖. 𝑒. 𝐸(𝑋𝑌) = 𝑀 (𝑡 , 𝑡 ) |
𝜕𝑡1 𝜕𝑡2 𝑋,𝑌 1 2 (𝑡1 , 𝑡2 ) = (0,0)
3.5.2 Theorem
If 𝑀𝑋,𝑌 (𝑡1 , 𝑡2 ) exists and X and Y are independent, then 𝑀𝑋,𝑌 (𝑡1 , 𝑡2 ) = 𝑀𝑋 (𝑡1 )𝑀𝑌 (𝑡2 ).

Proof
𝑀𝑋,𝑌 (𝑡1 , 𝑡2 ) = 𝐸(𝑒 𝑡1 𝑋+𝑡2 𝑌 )
= 𝐸(𝑒 𝑡1 𝑋 𝑒 𝑡2 𝑌 )
= 𝐸(𝑒 𝑡1 𝑋 ) 𝐸(𝑒 𝑡2 𝑌 ) 𝑏𝑦 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒
= 𝑀𝑋 (𝑡1 )𝑀𝑌 (𝑡2 )
Example
Random variables X and Y have joint pdf 𝑓(𝑥, 𝑦) = 2𝑒 −𝑥−𝑦 , 0 < 𝑥 < 𝑦 < ∞. Find
(a) the joint moment generating function of X and Y
(b) the marginal moment generating functions of X and Y.
(c) 𝜌(𝑋, 𝑌).
Soln

1
(a)
y
y=x

x
Remember the shaded region can be represented in two ways either by fixing x and letting y vary
or fixing y and letting x vary. If we fix y we have the following:
𝑀𝑋,𝑌 (𝑡1 , 𝑡2 ) = 𝐸(𝑒 𝑡1 𝑋+𝑡2 𝑌 )

= ∫ ∫ 𝑒 𝑡1 𝑥+𝑡2 𝑦 𝑓(𝑥, 𝑦)𝑑𝑥𝑑𝑦

𝑎𝑙𝑙 𝑦 𝑎𝑙𝑙 𝑥
∞ 𝑦

= ∫ ∫ 𝑒 𝑡1 𝑥+𝑡2 𝑦 2𝑒 −𝑥−𝑦 𝑑𝑥𝑑𝑦

𝑦=0 𝑥=0
∞ 𝑦

= ∫ ∫ 2𝑒 −𝑦+𝑡2 𝑦 𝑒 −𝑥+𝑡1 𝑥 𝑑𝑥𝑑𝑦

𝑦=0 𝑥=0
∞ 𝑦

= ∫ 2𝑒 −(1−𝑡2 )𝑦 ∫ 𝑒 −(1−𝑡1 )𝑥 𝑑𝑥𝑑𝑦

𝑦=0 𝑥=0
∞ 𝑦
−(1−𝑡2 )𝑦
𝑒 −(1−𝑡1 )𝑥
= ∫ 2𝑒 | 𝑑𝑦
−(1 − 𝑡1 )
𝑦=0 𝑥=0
∞
2
=− ∫ 𝑒 −(1−𝑡2 )𝑦 (𝑒 −(1−𝑡1 )𝑦 − 1)𝑑𝑦
1 − 𝑡1
𝑦=0
∞
2
=− ∫ (𝑒 −(2−𝑡1 −𝑡2 )𝑦 − 𝑒 −(1−𝑡2 )𝑦 )𝑑𝑦
1 − 𝑡1
𝑦=0
2 𝑒 −(2−𝑡1 −𝑡2 )𝑦 ∞ 𝑒 −(1−𝑡2 )𝑦 ∞
=− ( | + | )
1 − 𝑡1 −(2 − 𝑡1 − 𝑡2 ) 0 1 − 𝑡2 0
2 1 1
=− (− (0 − 1) + (0 − 1)) , 2 − 𝑡1 − 𝑡2 > 0 , 1 − 𝑡2 > 0
1 − 𝑡1 2 − 𝑡1 − 𝑡2 1 − 𝑡2
2 1 1
=− ( − ) , 2 > 𝑡1 + 𝑡2 , 1 > 𝑡2
1 − 𝑡1 2 − 𝑡1 − 𝑡2 1 − 𝑡2

2
 2 1 − 𝑡2 − 2 + 𝑡1 + 𝑡2
=− ( ) , 𝑡1 + 𝑡2 < 2 , 𝑡2 < 1
1 − 𝑡1 (2 − 𝑡1 − 𝑡2 )(1 − 𝑡2 )
2 −1 + 𝑡1
=− ( )
1 − 𝑡1 (2 − 𝑡1 − 𝑡2 )(1 − 𝑡2 )
2 1 − 𝑡1
= ( )
1 − 𝑡1 (2 − 𝑡1 − 𝑡2 )(1 − 𝑡2 )
2
= , 𝑡 + 𝑡2 < 2 , 𝑡2 < 1
(2 − 𝑡1 − 𝑡2 )(1 − 𝑡2 ) 1
(b)
𝑀𝑋 (𝑡1 ) = 𝑀𝑋 ,𝑌 (𝑡1 , 0)
2
= ,𝑡 < 2
(2 − 𝑡1 − 0)(1 − 0) 1
2
= ,𝑡 < 2
2 − 𝑡1 1
𝑂𝑅
2
𝑀𝑋 (𝑡) = ,𝑡 < 2
2−𝑡
𝑀𝑌 (𝑡2 ) = 𝑀𝑋 ,𝑌 (0, 𝑡2 )
2
= , 𝑡 < 2 , 𝑡2 < 1
(2 − 0 − 𝑡2 )(1 − 𝑡2 ) 2
2
= ,𝑡 < 1
(2 − 𝑡2 )(1 − 𝑡2 ) 2
𝑂𝑅
2
𝑀𝑌 (𝑡) = ,𝑡 < 1
(2 − 𝑡)(1 − 𝑡)
(c)
𝐶𝑜𝑣(𝑋, 𝑌)
𝜌(𝑋, 𝑌) =
√𝑉𝑎𝑟(𝑋)𝑉𝑎𝑟(𝑌)
𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸(𝑋𝑌) − 𝐸(𝑋)𝐸(𝑌)
𝐸(𝑋) = 𝑀𝑋′ (0)
𝑀𝑋 (𝑡) = 2(2 − 𝑡)−1
𝑀𝑋′ (𝑡) = −2(2 − 𝑡)−2 (−1) = 2(2 − 𝑡)−2
1
𝐸(𝑋) = 2(2 − 0)−2 = 2(2)−2 =
2
𝑀𝑋′′ (𝑡) = −2(2)(2 − 𝑡)−3 (−1) = 4(2 − 𝑡)−3
1
𝐸(𝑋 2 ) = 𝑀𝑋′′ (0) = 4(2 − 0)−3 = 4(2)−3 =
2
2 2
𝑀𝑌 (𝑡) = = 2
= 2(2 − 3𝑡 + 𝑡 2 )−1
(2 − 𝑡)(1 − 𝑡) 2 − 3𝑡 + 𝑡
2(3 − 2𝑡)
𝑀𝑌′ (𝑡) = −2(2 − 3𝑡 + 𝑡 2 )−2 (−3 + 2𝑡) =
(2 − 3𝑡 + 𝑡 2 )2

3
 2(3 − 0) 6 3
𝐸(𝑌) = 𝑀𝑌′ (0) = 2
= =
(2 − 0 + 0) 4 2
(2 − 3𝑡 + 𝑡 ) (−2) − (3 − 2𝑡)(2)(2 − 3𝑡 + 𝑡 2 )(−3 + 2𝑡)
2 2
𝑀𝑌′′ (𝑡) = 2( )
(2 − 3𝑡 + 𝑡 2 )4
−2(2 − 3𝑡 + 𝑡 2 )2 + 2(3 − 2𝑡)2 (2 − 3𝑡 + 𝑡 2 )
= 2( )
(2 − 3𝑡 + 𝑡 2 )4
′′ (0)
−2(2 − 0 + 0)2 + 2(3 − 0)2 (2 − 0 + 0) 28 7
𝐸(𝑌 2 ) = 𝑀𝑌 = 2( 4
)= =
(2 − 0 + 0) 8 2
2
𝜕
𝐸(𝑋𝑌) = 𝑀 (𝑡 , 𝑡 ) |
𝜕𝑡1 𝜕𝑡2 𝑋,𝑌 1 2 (𝑡1 , 𝑡2 ) = (0,0)
2 2
𝑀𝑋,𝑌 (𝑡1 , 𝑡2 ) = = (2 − 𝑡1 − 𝑡2 )−1
(2 − 𝑡1 − 𝑡2 )(1 − 𝑡2 ) (1 − 𝑡2 )
𝜕 2
𝑀 (𝑡 , 𝑡 ) = (−1)(2 − 𝑡1 − 𝑡2 )−2 (−1) = 2(1 − 𝑡2 )−1 (2 − 𝑡1 − 𝑡2 )−2
𝜕𝑡1 𝑋,𝑌 1 2 (1 − 𝑡2 )
𝜕
𝑀 (𝑡 , 𝑡 ) = 2[(2 − 𝑡1 − 𝑡2 )−2 (−1)(1 − 𝑡2 )−2 (−1) + (1 − 𝑡2 )−1 (−2)(2 − 𝑡1 − 𝑡2 )−3 (−1)]
𝜕𝑡2 𝜕𝑡1 𝑋,𝑌 1 2
= 2[(2 − 𝑡1 − 𝑡2 )−2 (1 − 𝑡2 )−2 + 2(1 − 𝑡2 )−1 (2 − 𝑡1 − 𝑡2 )−3 ]
𝐸(𝑋𝑌) = 2[2−2 (1) + 2(1)(2−3 )] = 2−1 + 2−1 = 1
1 3 1
𝐶𝑜𝑣(𝑋, 𝑌) =1− ( )=
2 2 4
2 1 1 2 1
𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − (𝐸(𝑋)) = − ( ) =
2 2 4
2
2 7 3 5
𝑉𝑎𝑟(𝑌) = 𝐸(𝑌 2 ) − (𝐸(𝑌)) = − ( ) =
2 2 4
1 1
4 1
𝜌(𝑋, 𝑌) = = 4 =
√5 √5
√1 (5)
4 4 4
= 0.4472

Exercise
Random variables X and Y have joint pdf 𝑓(𝑥, 𝑦) = 𝑒 −𝑥 , 0 < 𝑦 < 𝑥 < ∞. Find
(a) the joint moment generating function of X and Y.
(b) the marginal moment generating functions of X and Y.
(c) 𝜌(𝑋, 𝑌).