Review of Random Variables

1 Key functions and examples

 First remarks

• Reading Assignment: Review your Probability M362K materials:

- definition of discrete and continuous random variables;
- properties of random variables;
- functions associated with random variables . . .
• As inspiration, let us briefly discuss modeling; read Chapter 1. in the
textbook for more background
 Review of Random Variables

1 Key functions and examples

 Cumulative distribution function:
Definition

• Definition: The cumulative distribution function (also called the

distribution function) FX : R → [0, 1] of a random variable X is
defined as

FX (x) = P[X ≤ x], for x ∈ R

• Conventions:
We usually abbreviate “cumulative distribution function” to cdf
It is customary to label (in the right subscript) the cdf by the
random variable to which it “belongs”
 Cumulative distribution function:
Properties

Draw a graph!!!
• Image is [0, 1]
• Nondecreasing
• Right-continuous
•

limx→−∞ FX (x) = 0

limx→∞ FX (x) = 1
 Support: Definition

• Definition: The support of a random variable X is defined as the

set of numbers that are possible values of the random variable.
• Note: If you look at X as a function from the set of possible
outcomes to the real numbers, then its support is the image of X
 Discrete random variables

• Definition: A random variable is called discrete if its support

contains at most countably many values.
• Examples:
- the indicator r.v.: support is {0, 1};
- Bernoulli: support is {0, 1};
- Binomial: support finite;
- geometric, Poisson: support is N - infinite and countable ...
• Question: What does the graph of the cdf of a discrete random
variable look like?
 Example: Model 3 [number of claims]
• Let X be a random variable representing the number of claims on
one policy in one year. We want to set up a model for X , i.e., give
its distribution via its cdf FX .
• Evidently, X should be a discrete random variable
• Question: What is the support of X ?
• We assume that the possible values for the number of claims are
{0, 1, 2, 3, 4}. This is the support of X .
• One possible model is to set


 0 x <0

0.5 0≤x <1





0.75 1≤x <2
FX (x) = F3 (x) =


 0.87 2≤x <3
0.95 3≤x <4





1 4≤x

 Continuous random variables

• Definition: A random variable is called continuous if its distribution

function is
(a) continuous, and
(b) differentiable everywhere with the possible exceptions of at most
countably many values.
 Example: Model 1 [age at death]
• Let X be a random variable representing the age at death of an
individual. We want to set up a model for X , i.e., give its
distribution via its cdf FX .
• Question: What is the support of X ?
• Evidently X ≥ 0.
So,

FX (x) = 0 for x < 0

• The maximal age is trickier .... We can agree on a certain maximal

age, say 100. So,

FX (x) = 1 for x > 100

• For x ∈ [0, 100], one possibility is to assume that any age at death is
equally likely. So, FX is linear on that interval, i.e.,
1
FX (x) = 100 x for 0 ≤ x ≤ 100

• Can you think of other “plausible” models for the age at death?
 Survival function: Definition

• Definition: The survival function SX : R → [0, 1] of a random

variable X is defined as

SX (x) = 1 − FX (x) = P[X > x], for x ∈ R

• Conventions:
It is customary to label (in the right subscript) the survival function
by the random variable to which it “belongs”
 Survival function: Properties

• Image is [0, 1]
• Nonincreasing
• Right-continuous
•

limx→−∞ SX (x) = 1

limx→∞ SX (x) = 0
 Example: Model 1 [age at death] - cont’d

• The survival function for the above model is


1
 for x < 0
SX (x) = 100−x100 for 0 ≤ x ≤ 100

0 for x > 100

 Mixed random variables

• Definition: A random variable is called mixed if

(a) it is not discrete, and
(b) its cumulative distribution function is continuous everywhere except
for at least one and at most countably many points, and
(c) its cumulative distribution function is differentiable everywhere
except for at most countably many points.
 Example: Model 4 [benefit payments]

• Let X represent the total dollars paid on a policy in one year.

• Clearly, its support is contained in [0, ∞)
• One possible model is to set
(
0 x <0
FX (x) = F4 (x) = −5
1 − 0.3e −10 x x ≥0
 The probability density function

• Definition: The probability density function (also called the density

function) fX : R → R+ of a random variable X is defined as

fX (x) = FX0 (x) = −SX0 (x),

for every x ∈ R at which the derivative FX0 (x) exists

• Convention: The abbreviation pdf is common
• Only the densities of continuous or mixed random variables “make
sense”; the density of a discrete random variable is not defined
• When fX is defined over an interval (a, b), the Fundamental
Theorem of Calculus yields:
Z b
P[a < X ≤ b] = FX (b) − FX (a) = fX (x) dx
a
 The probability mass function

• Definition: The probability function (also called the probability

mass function) pX : R → [0, 1] of a random variable X is given by

pX (x) = P[X = x] for x ∈ R

• Convention: The abbreviation pmf is common.

• The pmf only “makes sense” for discrete or mixed random variables;
for continuous random variables we can say that it is not defined
• For discrete r.v.s, we have FX (x) = y ≤x pX (y ) and
P
P
SX (x) = y >x pX (y )
 Example: Model 4 - cont’d
• Let X represent the total dollars paid on a policy in one year with
(
0 x <0
FX (x) = F4 (x) = −5
1 − 0.3e −10 x x ≥0
• This is a mixed r.v.
• The event {X = 0} has a positive probability:

P[X = 0] = F4 (0) − F4 (0−) = 0.7 − 0 = 0.7

• At all x 6= 0, F4 is differentiable, and we have
(
0 0 x <0
f4 (x) = F4 (x) = −5
3 × 10−6 e −10 x x >0
• Throughout this course, we will abuse the notation slightly and write
(
0.7 x =0
f4 (x) = −6 −10−5 x
3 × 10 e x >0
understanding that f4 (x) = 0 otherwise and that the value assigned
to f4 (0) is not really the density
 The Hazard Rate

• Definition: The hazard rate (also known as the force of mortality

and the failure rate) of a r.v. X is a function hX : I → R+ is defined
as
fX (x) S 0 (x)
hX (x) = SX (x) = − SXx (x) = −d[ln(SX (x))]

for all x ∈ I , where I is the set of all real numbers at which the
density fX is defined and SX 6= 0.
• The terminology depends on the context.
• To recover the survival function from the hazard rate we use
Z b
SX (x) = exp{− h(x) dx}
0
 The Mode

• Definition: The mode of a discrete r.v. X is the value (or a set of

values) that has the largest probability of occurring, i.e.,

mode of X = argmax pX (x)

x

The modes of a continuous r.v. X are the local maxima of its

density function.
 Final remarks

• You are now responsible for Chapter 2. from the textbook.

• Read the material in it,
go over the given examples with particular care to the Models we did
not cover in class, and
(at least) attempt the Exercises in the end of the chapter.
• Definitions and basic properties of distributions “with names” that
will be studied in this course can be found in Appendices A and B;
start getting (re)acquainted with them.