Scribd
Upload
5 views

Lecture 17

The document discusses transformations in probability, specifically deriving joint densities from given joint densities using Jacobians. It provides examples involving exponential and normal random variables, detailing the process of finding densities for transformed variables. The document emphasizes the importance of the Jacobian determinant and offers alternative methods for calculating densities.

Uploaded by

Thành Nguyễn
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
5 views

Lecture 17

The document discusses transformations in probability, specifically deriving joint densities from given joint densities using Jacobians. It provides examples involving exponential and normal random variables, detailing the process of finding densities for transformed variables. The document emphasizes the importance of the Jacobian determinant and offers alternative methods for calculating densities.

Uploaded by

Thành Nguyễn
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Lecture 17: Transformations

STOR 435, Spring, 2025

3/31/2025

435-Spring-2025 transformation
Basic setting and recipe

Pitman: Section 5.4


Setting: Consider the joint density f(X,Y) (x, y), (x, y) ∈ A and
the map (x, y) → (u, v). Our objective is to derive the joint
density f(U,V) (u, v), (u, v) ∈ D.
Recipe: Compute the Jacobian factor J, i.e. the absolute
∂x ∂x
∂u ∂v
value of the determinant , then we have the
∂y ∂y
∂u ∂v
expression

f(U,V) (u, v) = J · f(X,Y) (x(u, v), y(u, v)) .

Note: An alternative way to calculate J will be given in


Example 2.

435-Spring-2025 transformation
Example 1

Note: Special care is required in performing multiple


integration with some variables defined over a restricted
domain.
Example 1: Let X and Y be iid Exp(λ) RV’s. Define
U = X + Y and V = X − Y. Derive the densities f(U,V) (u, v),
fU (u) and fV (v).
Restrictions: {x > 0, y > 0} is transformed to
{u > −v, u > v} or {u > max{−v, v}} or {u > |v|}.
The Jacobian determinant for (u, v) → (x, y) has its
absolute value 1/2. Therefore,

1 λ2 −λu
f(U,V) (u, v) = f(X,Y) ((u + v)/2, (u − v)/2) = e , u > |v|.
2 2

435-Spring-2025 transformation
Example 1 continued

Furthermore,

fU (u) = λ2 u e−λu , u > 0. (U ∼ Gamma(2, λ))


Z ∞
λ
fV (v) = f(U,V) (u, v) du = e−λ|v| , v ∈ IR.
|v| 2
V follows a double exponential distribution with parameter
λ > 0.

435-Spring-2025 transformation
Example 2
Let X and Y be iid N(0, 1) RV’s. Define U = XY and V = X/Y.
Note: UV = X 2 and U/V = Y 2 hence we have the
restriction D ≜ {u > 0, v > 0} ∪ {u < 0, v < 0}.
The Jacobian determinant for (u, v) → (x, y) has its
1
absolute value 2|v| . Hence

1
f(U,V) (u, v) = e−u(v+1/v)/2 , (u, v) ∈ D.
4π |v|

and
Z
1
fV (v) = f(U,V) (u, v) du = · · · = , v ∈ IR.
RI π (v2 + 1)

V follows a Cauchy distribution.


Note: fU (u) does not enjoy a closed form expression.

435-Spring-2025 transformation
Example 2 continued
If we are only interested in V = X/Y and its density fV (v), but
not U = XY, the following alternative method works better.
Step 1: Let U = X (easy to work with) along with V = X/Y.
Then the Jacobian determinant for (u, v) → (x, y) has its
absolute value |u|
v2
.
Step 2: The joint density

|u| |u| u2 (1+v−2 )/2


f(U,V) (u, v) = f (u, u/v) = e , (u, v) ∈ IR2 .
v2 (X,Y) 2πv2
Step 3: The marginal density
Z
1
fV (v) = f (U, V)(u, v) du = · · · = , v ∈ IR.
RI π (v2+ 1)

Note: U serves as an auxiliary variable which can be


defined arbitrarily as long as it is easy to integrate it out.

435-Spring-2025 transformation

You might also like