34

Definition 1.125 (Probability Mass Function (p.m.f.)). Let X be a discrete RV with DF FX and
support S. Consider the function fX : R æ R defined by
Y
_
X (x) ≠ FX (x≠) = P(X = x), if x œ S,
_
]F
fX (x) := _
[0, if x œ S c .
_

This function fX is called the probability mass function (p.m.f.) of X.

Example 1.126. Continuing with the Example 1.124, the p.m.f. fX is given by
Y
1
_
_
_
_ 4
, if x = 0,
_
_
_
_
]1,
_ if x = 1,
2
fX (x) =
_ 1
_
_
_
_ 4
, if x = 2.,
_
_
_
_
[ 0, otherwise.

Remark 1.127. Let X be a discrete RV with DF FX , p.m.f. fX and support S. Then we have the
following observations.

(a) Continuing the discussion from Remark 1.122, we have for all A ™ R,
ÿ
PX (A) = P(X œ A) = fX (x).
xœAflS

(b) As a special case of the previous observation, note that for A = (≠Œ, x], x œ R, we obtain
ÿ
FX (x) = P(X Æ x) = P(X œ (≠Œ, x]) = fX (t).
tœ(≠Œ,x]flS

Therefore, the p.m.f. fX is uniquely determined by the DF FX and vice versa.

(c) To study a discrete RV X, we may study any one of the following three quantities, viz. the
law/distribution PX , the DF FX or the p.m.f. fX . Given any one of these quantities, the
other two can be obtained using the relations described above.
 35

(d) By Definition 1.121 and Definition 1.125, we have that the p.m.f. fX : R æ R is a function
such that
ÿ
fX (x) = 0, ’x œ S c , fX (x) > 0, ’x œ S, fX (x) = 1.
xœS

Remark 1.128. Let ÿ =

” S µ R be a finite or countably infinite set and let f : R æ R be such that
ÿ
f (x) = 0, ’x œ S c , f (x) > 0, ’x œ S, f (x) = 1.
xœS

Then by an argument similar to Proposition 1.45, we conclude that P as defined below is a prob-
ability function/measure on B, where B denotes the power set of R. For all A ™ R, consider
ÿ
P(A) := f (x).
xœAflS

By an argument similar to Theorem 1.115, we can then show that the function F : R æ R defined
by F (x) := P((≠Œ, x]), ’x œ R is non-decreasing, right continuous with limxæ≠Œ F (x) = 0 and
limxæŒ F (x) = 1. By Theorem 1.116, this F is the DF of some RV Y , i.e. FY = F and by
construction, Y must be discrete with support S and p.m.f. fY = f .

Example 1.129. Take S to be the set of natural numbers {1, 2, · · · } and consider the function
f : R æ R defined by Y
_
]1
_ , if x œ S,
2x
f (x) :=
_
[0,
_ if x œ S c .
q qΠ1
Then f takes non-negative values with xœS f (x) = n=1 2n = 1. Therefore f is the p.m.f. of
some RV X with DF FX given by
ÿ
FX (x) = P(X Æ x) = fX (t)
tœ(≠Œ,x]flS
Y Y
_ _
]0,
_ if x < 1, ]0,
_ if x < 1,
= _q =_
1 [1 ≠ 1m , if x œ [m, m + 1), m œ S.
_
[ m
n=1 2n , if x œ [m, m + 1), m œ S. _
2
36

Definition 1.130 (Continuous RV and its Probability Density Function (p.d.f.)). An RV X is

said to be a continuous RV if there exists an integrable function f : R æ [0, Œ) such that
⁄ x
FX (x) = P(X Æ x) = f (t) dt, ’x œ R.
≠Œ

The function f is called the probability density function (p.d.f.) of X.

Remark 1.131. Let X be a continuous RV with DF FX and p.d.f. fX . Then we have the following
observations.
sx
(a) Since fX is integrable, from the relation FX (x) = ≠Œ fX (t) dt, ’x œ R, we have FX is
continuous on R. In particular, FX is absolutely continuous. Moreover, for all a < b, we
have ⁄ b ⁄ a ⁄ b
FX (b) ≠ FX (a) = fX (t) dt ≠ fX (t) dt = fX (t) dt.
≠Œ ≠Œ a
(b) Since FX is continuous, we have
(i) FX (x≠) = FX (x) = FX (x+), ’x œ R.
(ii) P(X = x) = PX ({x}) = FX (x) ≠ FX (x≠) = 0, ’x œ R.
(iii) P(X < x) = FX (x≠) = FX (x) = P(X Æ x), ’x œ R.
(iv) For all a < b,

P(a < X < b) = P(a < X Æ b) = P(a Æ X < b) = P(a Æ X Æ b)

⁄ b
= FX (b) ≠ FX (a) = fX (t) dt.
a

(c) If A µ R is finite or countably infinite, then by the finite/countable additivity of PX , we

have
ÿ
P(X œ A) = PX (A) = PX ({x}) = 0.
xœA

(d) By definition, we have fX (x) Ø 0, ’x œ R and

⁄ x ⁄ Œ
1 = lim FX (x) = lim fX (t) dt = fX (t) dt.
xæŒ xæŒ ≠Œ ≠Œ
sŒ
Remark 1.132. Let f : R æ [0, Œ) be an integrable function with ≠Œ f (t) dt = 1. Then the
sx
function F : R æ [0, 1] defined by F (x) := ≠Œ f (t) dt, ’x œ R is non-decreasing and continuous
 37

with limxæ≠Œ F (x) = 0 and limxæŒ F (x) = 1. By Theorem 1.116, this F is the DF of some RV
Y , i.e. FY = F and by construction, Y must be continuous with p.d.f. fY = f .

Example 1.133. Let X be an RV with the DF FX : R æ R as discussed in Example 1.118. Here,

Y
_
_
_
_0, if x < 0,
_
]
FX (x) := x, if 0 Æ x < 1,
_
_
_
_
_
[1, if x Ø 1.

Then the function f : R æ [0, Œ) defined by

Y
_
]1,
_ if 0 Æ x Æ 1,
f (x) :=
_
[0,
_ otherwise
sx
is an integrable function with FX (x) = ≠Œ f (t) dt, ’x œ R. Therefore, X is a continuous RV with
p.d.f. f .

Example 1.134. Consider the DF F : R æ [0, 1] considered in Example 1.120 given by

Y
_
_
_
_0, if x < 0,
_
_
_
_
]1 + x,
_ if 0 Æ x Æ 1,
4 2
F (x) =
_ 1
_
_
_
_ 2
+ x4 , if 1 < x < 2,
_
_
_
_
[ 1, if x Ø 2.

As discussed earlier, F has a discontinuity at the point 0. Therefore, an RV X with DF F is not

a continuous RV.

Note 1.135. Given a continuous RV X with p.d.f. fX , the DF FX is computed by the formula
sx
FX (x) = ≠Œ fX (t) dt, ’x œ R.
38

Example 1.136. Consider a function f : R æ R of the form

Y
_
_
_
_–x, if x œ [≠1, 0),
_
]
f (x) = x2
8
, if x œ [0, 2],
_
_
_
_
_
[ 0, otherwise

for some – œ R. For this f to be a p.d.f. of a continuous RV, two conditions need to be satisfied,
sŒ
viz. f (x) Ø 0, ’x œ R and ≠Œ f (x) dx = 1.
The first condition is satisfied on (≠Œ, ≠1) fi [0, Œ). For x œ [≠1, 0), we must have –x Ø 0, which
implies – Æ 0.
s0 s 2 x2
From the second condition, we have ≠1 –x dx + 0 8 dx = 1. This yields – = ≠ 43 , which satisfies
– Æ 0.
Therefore, for f to be a p.d.f. we must have – = ≠ 43 .

In what follows, we consider the question of computing fX from the DF FX .

Remark 1.137 (Is the p.d.f. of a continuous RV unique?). Let X be a continuous RV with DF FX
and p.d.f. fX . Fix any finite or countably infinite set A µ R and fix c Ø 0. Consider the function
g : R æ [0, Œ) defined by Y
_
X (x), if x œ Ac ,
_
]f
g(x) :=
_
_
[c, if x œ A.
sx
Then g is integrable and FX (x) = ≠Œ g(t) dt, ’x œ R. Hence, g is also a p.d.f. for X. Therefore,
the RV X with DF FX is a continuous RV with p.d.f. f (or g). For example,
Y
_
]1,
_ if 0 < x < 1,
g(x) :=
_
[0,
_ otherwise

is a p.d.f. for X as in Example 1.133. More generally, we may also consider

Y
_
X (x), if x œ Ac ,
_
]f
g(x) :=
_
_
[cx , if x œ A
 39

as a p.d.f., where cx Ø 0, ’x œ A.

Note 1.138. In fact, a p.d.f. fX for a continuous RV X is determined uniquely on the complement
of sets of ‘length 0’, such as sets which are finite or countably infinite. We do not make a precise
statement – this is beyond the scope of this course. However, we consider the deduction of p.d.f.s
from the DFs.

The next result is stated without proof.

Theorem 1.139. Let X be an RV with DF FX .

sŒ
(a) If FX is differentiable on R with ≠Œ FXÕ (t) dt = 1, then X is a continuous RV with p.d.f.
FXÕ .
(b) If FX is differentiable everywhere except on a finite or a countably infinite set A µ R with
sŒ
≠Œ FXÕ (t) dt = 1, then X is a continuous RV with p.d.f. f given by
Y
_
X (x), if x œ Ac ,
_
]F Õ
f (x) :=
_
[0,
_ if x œ A.
Note 1.140. Continuing the discussion from Note 1.135, the DF FX of a continuous RV X may
be used to compute the p.d.f. fX . In Example 1.133, the DF FX is given by
Y
_
_
_
_0, if x < 0,
_
]
FX (x) := _x, if 0 Æ x < 1,
_
_
_
_
[1, if x Ø 1.

It is differentiable everywhere except at the points 0 and 1. Using Theorem 1.139, we have the
p.d.f. given by Y
_
]1,
_ if 0 < x < 1,
f (x) := _
[0, otherwise.
_

Note 1.141. To study a continuous RV X, we may study any one of the following three quantities,
viz. the law/distribution PX , the DF FX or the p.d.f. fX . Given any one of these quantities, the
other two can be obtained using the relations described above.
40

Definition 1.142 (Support of a Continuous RV). Let X be a continuous RV with DF FX . The

set
S := {x œ R : FX (x + h) ≠ FX (x ≠ h) > 0, ’h > 0}

is defined to be the support of X.

Remark 1.143. The support S of a continuous RV X can be expressed in terms of the law/distribution
of X as follows.

S = {x œ R : P(x ≠ h < X Æ x + h) > 0, ’h > 0} = {x œ R : PX ((x ≠ h, x + h]) > 0, ’h > 0}.

Remark 1.144. The support S of a continuous RV X can be expressed in terms of the p.d.f. fX as
follows. ⁄ x+h
S = {x œ R : fX (t) dt > 0, ’h > 0}.
x≠h

Note 1.145. If x œ
/ S, where S is the support of a continuous RV X, then there exists h > 0 such
that FX (x + h) = FX (x ≠ h). By the non-decreasing property of FX , we conclude that FX remains
a constant on the interval [x ≠ h, x + h]. In particular, fX (t) = FXÕ (t) = 0, ’t œ (x ≠ h, x + h).

Example 1.146. Consider a continuous RV X with DF FX : R æ [0, 1] and p.d.f. fX : R æ [0, Œ)

given by Y
_
_
_
_0, if x < 0, Y
_
_
] ]1,
_ if 0 < x < 1,
FX (x) := fX (x) :=
_x, if 0 Æ x < 1,
,
_
_
_
_ [0,
_ otherwise.
_
[1, if x Ø 1.
To identify the support S, we consider the following cases.
(a) Let x œ (≠Œ, 0). Then for all h with ≠x > h > 0, we have x ≠ h < x + h < 0 and
consequently, FX (x + h) ≠ FX (x ≠ h) = 0 ≠ 0 = 0. Therefore x œ
/ S.
(b) Let x œ (1, Œ). Then for all 0 < h < x ≠ 1, we have 1 < x ≠ h < x + h and consequently,
FX (x + h) ≠ FX (x ≠ h) = 1 ≠ 1 = 0. Therefore x œ
/ S.
(c) Let x œ (0, 1). For any 0 < h < min{x, 1 ≠ x}, we have 0 < x ≠ h < x + h < 1 and
consequently, FX (x + h) ≠ FX (x ≠ h) = (x + h) ≠ (x ≠ h) = 2h > 0. For h Ø min{x, 1 ≠ x},
 41

at least one of x ≠ h, x + h is in (0, 1)c and hence FX (x + h) ≠ FX (x ≠ h) > 0. Therefore

x œ S.
(d) Let x = 0. Then for any h > 0, we have FX (0 + h) ≠ FX (0 ≠ h) = FX (0 + h) > 0. Then
0 œ S. By a similar argument, 1 œ S.

From the above discussion, we conclude that S = [0, 1].

Remark 1.147 (Identifying discrete/continuous RVs from their DFs). Suppose that the distribution
of an RV X is specified by a given DF FX . In order to check if X is a discrete/continuous RV, we
use the following steps.

(a) Identify the set D = {x œ R : FX (x≠) < FX (x+)} = {x œ R : P(X = x) > 0} of

discontinuities of FX . Recall that D is a finite or a countably infinite set.
(b) If D is empty, then FX is continuous on R. By verifying the hypothesis of Theorem 1.139
or otherwise, check if there exists a p.d.f.. If a p.d.f. exists, then X is a continuous RV.
Otherwise, X is not a continuous RV.
(c) If FX has at least one discontinuity, then FX is not continuous on R and hence X cannot
be a continuous RV. For X to be a discrete RV X, we must have
ÿ ÿ
[FX (x+) ≠ FX (x≠)] = P(X = x) = 1.
xœD xœD

If the above condition is satisfied, X is a discrete RV. Otherwise, X is not a discrete RV.

Note 1.148. Cantor function (also known as the Devil’s Staircase) is an example of a continuous
distribution function, which is not absolutely continuous. In this case, the DF F is not repre-
sx
sentable as ≠Œ f (t) dt for any non-negative integrable function. We do not discuss these types of
examples in this course.
42

Note 1.149. Consider the DF F : R æ [0, 1] considered in Example 1.120 given by

Y
_
_
_
_0, if x < 0,
_
_
_
_
]1 + x,
_ if 0 Æ x Æ 1,
4 2
F (x) =
_ 1
_
_
_
_ 2
+ x4 , if 1 < x < 2,
_
_
_
_
[1, if x Ø 2.

As discussed in Example 1.123 and Example 1.134, an RV with DF F is neither discrete nor
continuous.

Definition 1.150 (Quantiles and Median for an RV). Let X be an RV with DF FX . For any
p œ (0, 1), a number x œ R is called a quantile of order p if the following inequalities are satisfied,
viz.
p Æ FX (x) Æ p + P(X = x).
1
A quantile of order 2
is called a median.

Note 1.151. A quantile need not be unique. Refer to problem set 4 for explicit examples.

Notation 1.152. We write zp (X) to denote a quantile of order p.

1 3
Notation 1.153. The quantiles of order 4
and 4
for an RV X are referred to as the lower and
upper quartiles of X, respectively.

Note 1.154. The inequalities mentioned in Definition 1.150 can be restated as

P(X Æ x) Ø p, P(X Ø x) Ø 1 ≠ p.

Note 1.155. Let X be a continuous RV with DF FX . Then a quantile of order p is a solution to

the equation FX (x) = p, since P(X = x) = 0, ’x œ R. Moreover, if FX is strictly increasing, then
zp (X) is unique for all p œ (0, 1).