Lecture-4

Binomial r.v Approximations and Conditional

Probability Density Functions
Let X represent a Binomial r.v as in (3.42). Then from (2.32)
k2 k2
 n  k n −k
P (k1 ≤ X ≤ k 2 ) = ∑ Pn ( k ) = ∑   p q . (1)
k = k1 k = k1  k 

Since the binomial coefficient  nk  = ( n −nk! )! k ! grows quite

rapidly with n, it is difficult to compute (1) for large n. In
this context, two approximations are extremely useful.

a. The Normal Approximation (Demoivre-Laplace

Theorem) Suppose n → ∞ with p held fixed. Then for k in
the npq neighborhood of np, we can approximate 1
  n  k n−k 1 − ( k − np ) 2 / 2 npq
  p q ≈ e . (2)
k  2 π npq
Thus if k1 and k 2 in (1) are within or around the
neighborhood of the interval (np − npq , np + npq ), we can
approximate the summation in (1) by an integration. In that
case (1) reduces to
P (k1 ≤ X ≤ k2 ) = ∫
1 1 − y2 / 2
dx = ∫
k2 x2
−( x −np ) 2 / 2 npq
e e dy , (3)
k1 2πnpq x1 2π
where
k 1 − np k 2 − np
x1 = , x2 = .
npq npq

We can express (3) in terms of the normalized integral

1 x
∫
− y2 / 2
erf ( x ) = e dy = erf ( − x ) (4)
2π 0

that has been tabulated extensively (See Table 4.1). 2

For example, if x1 and x2 are both positive ,we obtain
P (k 1 ≤ X ≤ k 2 ) = erf ( x 2 ) − erf ( x1 ). (5)

Example 4.1: A fair coin is tossed 5000 times. Find the

probability that the number of heads is between 2475 to
2525.
Solution: We need P ( 2475 ≤ X ≤ 2525 ). Here n is large so
that we can use the normal approximation. In this case p = 12 ,

so that np = 2500 and npq ≈ 35 . Since np − npq = 2465,

and np + npq = 2535, the approximation is valid for k1 = 2475
and k2 = 2525 . Thus
P (k 1 ≤ X ≤ k 2 ) =
x2 1 − y2 /2
∫x1 2π
e dy .
k 1 − np 5 k − np 5
Here x1 = = − , x2 = 2 = .
npq 7 npq 7
3
 1 x 1
∫
− y2 / 2
erf ( x ) = e dy = G ( x ) −
2π 0 2

x erf(x) x erf(x) x erf(x) x erf(x)

0.05 0.01994 0.80 0.28814 1.55 0.43943 2.30 0.48928

0.10 0.03983 0.85 0.30234 1.60 0.44520 2.35 0.49061
0.15 0.05962 0.90 0.31594 1.65 0.45053 2.40 0.49180
0.20 0.07926 0.95 0.32894 1.70 0.45543 2.45 0.49286
0.25 0.09871 1.00 0.34134 1.75 0.45994 2.50 0.49379

0.30 0.11791 1.05 0.35314 1.80 0.46407 2.55 0.49461

0.35 0.13683 1.10 0.36433 1.85 0.46784 2.60 0.49534
0.40 0.15542 1.15 0.37493 1.90 0.47128 2.65 0.49597
0.45 0.17364 1.20 0.38493 1.95 0.47441 2.70 0.49653
0.50 0.19146 1.25 0.39435 2.00 0.47726 2.75 0.49702

0.55 0.20884 1.30 0.40320 2.05 0.47982 2.80 0.49744

0.60 0.22575 1.35 0.41149 2.10 0.48214 2.85 0.49781
0.65 0.24215 1.40 0.41924 2.15 0.48422 2.90 0.49813
0.70 0.25804 1.45 0.42647 2.20 0.48610 2.95 0.49841
0.75 0.27337 1.50 0.43319 2.25 0.48778 3.00 0.49865

Table 4.1
4
Since x1 < 0, from Fig. 4.1(b), the above probability is given
by P (2475 ≤ X ≤ 2525 ) = erf ( x 2 ) − erf ( x1 ) = erf ( x 2 ) + erf (| x1 |)
5
= 2 erf   = 0 . 516 ,
7
where we have used Table 4.1 (erf (0.7) = 0.258).
1 1
e−x
2
e−x
2 /2
/2

2π 2π

x x
x1 x2 x1 x2
(a) x1 > 0, x2 > 0 (b) x1 < 0, x2 > 0
Fig. 4.1
b. The Poisson Approximation
As we have mentioned earlier, for large n, the Gaussian
approximation of a binomial r.v is valid only if p is fixed,
i.e., only if np >> 1 and npq >> 1 . what if np is small, or if it
does not increase with n? 5
Obviously that is the case if, for example, p → 0 as n → ∞,
such that np = λ is a fixed number.
Many random phenomena in nature in fact follow this
pattern. Total number of calls on a telephone line, claims in
an insurance company etc. tend to follow this type of
behavior. Consider random arrivals such as telephone calls
over a line. Let n represent the total number of calls in the
interval (0, T ). From our experience, as T → ∞ we have n → ∞
so that we may assume n = µT . Consider a small interval of
duration ∆ as in Fig. 4.2. Had there been only a single call
coming in, the probability p of that single call occurring in
that interval must depend on its relative size with respect to
T. 1 2
 n

∆
0 T 6
Fig. 4.2
 ∆
Hence we may assume Note that p → 0 as T → ∞ .
p =
T
.
∆
However in this case np = µ T ⋅ = µ ∆ = λ is a constant,
T
and the normal approximation is invalid here.
Suppose the interval ∆ in Fig. 4.2 is of interest to us. A call
inside that interval is a “success” (H), whereas one outside is
a “failure” (T ). This is equivalent to the coin tossing
situation, and hence the probability Pn ( k ) of obtaining k
calls (in any order) in an interval of duration ∆ is given by
the binomial p.m.f. Thus
n!
Pn ( k ) = p k (1 − p ) n − k , (6)
( n − k )! k !

and here as n → ∞ , p → 0 such that np = λ . It is easy to

obtain an excellent approximation to (6) in that situation. To
see this, rewrite (6) as 7
 Pn ( k ) =
n ( n − 1)  (n − k + 1) (np ) k
(1 − np / n ) n − k
nk k!
(7)


1 
n 
2
=  1 − 1 − 
n
 
1 −

k − 1  λk (1 − λ / n ) n

n  k ! (1 − λ / n ) k
.

Thus λk − λ
lim Pn ( k ) = e , (8)
n → ∞ , p → 0 , np = λ k!
since the finite products 
1 −

1 
1 −
n 
2 

n 
  1 − k n− 1  as well
λ 
as  n  tend to unity as n → ∞ , and
k

 1 −

λ 
n

lim  1 −  = e − λ .
n→ ∞
 n 
The right side of (8) represents the Poisson p.m.f and the
Poisson approximation to the binomial r.v is valid in
situations where the binomial r.v parameters n and p
diverge to two extremes ( n → ∞ , p → 0 ) such that their
product np is a constant. 8
Example 4.2: Winning a Lottery: Suppose one million
lottery tickets are issued with 100 winning tickets among
them. (a) If a person purchases 100 tickets, what is the
probability of winning? (b) How many tickets should one
buy to be 95% confident of having a winning ticket?
Solution: The probability of buying a winning ticket
No. of winning tickets 100
p= = 6 = 10− 4.
Total no. of tickets 10
Here n = 100 , and the number of winning tickets X in the n
purchased tickets has an approximate Poisson distribution
with parameter λ = np = 100 × 10 − 4 = 10 − 2 . Thus
−λ λk
P( X = k ) = e ,
k!
and (a) Probability of winning = P( X ≥ 1) = 1 − P( X = 0) = 1 − e−λ ≈ 0.01.
9
(b) In this case we need P ( X ≥ 1) ≥ 0.95 .

P ( X ≥ 1) = 1 − e − λ ≥ 0 .95 implies λ ≥ ln 20 = 3.

But λ = np = n × 10 −4 ≥ 3 or n ≥ 30,000 . Thus one needs to buy

about 30,000 tickets to be 95% confident of having a
winning ticket!
Example 4.3: A space craft has 20,000 components (n → ∞ ).
The probability of any one component being defective
is 10 − 4 ( p → 0 ). The mission will be in danger if five or
more components become defective. Find the probability of
such an event.
Solution: Here n is large and p is small, and hence Poisson
approximation is valid. Thus np = λ = 20 , 000 × 10 − 4 = 2 ,
and the desired probability is given by
10
 4
λk 4
λk
P( X ≥ 5) = 1 − P( X ≤ 4) = 1 − ∑ e −λ
= 1− e ∑
−2

k =0 k! k =0 k!

 4 2
= 1 − e −2 1 + 2 + 2 + +  = 0.052.
 3 3

Conditional Probability Density Function

For any two events A and B, we have defined the conditional
probability of A given B as
P( A ∩ B ) (9)
P( A | B ) = , P( B ) ≠ 0.
P( B )
Noting that the probability distribution function F X ( x ) is
given by
F X ( x ) = P { X ( ξ ) ≤ x }, (10)
we may define the conditional distribution of the r.v X given
the event B as
11
 P { (X ( ξ ) ≤ x ) ∩ B }.
F X ( x | B ) = P { X ( ξ ) ≤ x | B }= (11)
P(B)
Thus the definition of the conditional distribution depends
on conditional probability, and since it obeys all probability
axioms, it follows that the conditional distribution has the
same properties as any distribution function. In particular
P { (X ( ξ ) ≤ +∞ )∩ B }=
P(B)
F X ( +∞ | B ) = = 1,
P(B ) P(B)
(12)
P { (X ( ξ ) ≤ −∞ )∩ B } = P (φ ) = 0 .
F X −∞ B =
( | )
P(B ) P(B)
Further
P { (x 1 < X (ξ ) ≤ x 2 ) ∩ B }
P ( x 1 < X (ξ ) ≤ x 2 | B ) =
P(B)
= F X ( x 2 | B ) − F X ( x 1 | B ), (13)
12
Since for x2 ≥ x1,
(X (ξ ) ≤ x 2 ) = (X (ξ ) ≤ x1 ) ∪ (x1 < X (ξ ) ≤ x 2 ). (14)

The conditional density function is the derivative of the

conditional distribution function. Thus
dFX ( x | B )
f X ( x | B) = , (15)
dx
and proceeding as in (3.26) we obtain
x
FX ( x | B ) = ∫ −∞
f X ( u | B ) du . (16)

Using (16), we can also rewrite (13) as

P (x1 < X (ξ ) ≤ x 2 | B ) =
x2
∫ x1
f X ( x | B ) dx . (17)

13
Example 4.4: Refer to example 3.2. Toss a coin and X(T)=0,
X(H)=1. Suppose B = {H }. Determine FX ( x | B ).
Solution: From Example 3.2, FX ( x ) has the following form.
We need FX ( x | B ) for all x.
For x < 0, {X (ξ ) ≤ x }= φ , so that { (X (ξ ) ≤ x ) ∩ B }= φ ,
and FX ( x | B ) = 0.

FX (x) FX (x)

1 1
q
x x
1 1
(a) (b)
Fig. 4.3 14
For 0 ≤ x < 1, {X (ξ ) ≤ x }= {T }, so that
{ (X (ξ ) ≤ x ) ∩ B }= {T }∩ { H }= φ and FX ( x | B ) = 0.
For x ≥ 1, {X (ξ ) ≤ x }= Ω , and
P( B )
{ (X (ξ ) ≤ x ) ∩ B }= Ω ∩ { B }= { B } and FX ( x | B) = =1
P( B )
(see Fig. 4.3(b)).
Example 4.5: Given FX ( x ), suppose B = {X (ξ ) ≤ a}. Find f X ( x | B ).
Solution: We will first determine FX ( x | B ). From (11) and B
as given above, we have
P { (X ≤ x ) ∩ (X ≤ a ) }
FX ( x | B ) = . (18)
P (X ≤ a )

15
For x < a , (X ≤ x ) ∩ (X ≤ a ) = (X ≤ x ) so that
P (X ≤ x )= FX (x)
FX (x | B ) = . (19)
P (X ≤ a ) FX (a )
For x ≥ a , (X ≤ x ) ∩ (X ≤ a ) = ( X ≤ a ) so that FX ( x | B ) = 1.
Thus
 FX ( x )
 , x < a,
FX ( x | B ) =  FX (a ) (20)
 1, x ≥ a,

and hence
 fX (x)
d  , x < a,
fX (x | B) = FX ( x | B ) =  FX (a ) (21)
dx  0 , otherwise.

16
 FX ( x | B )
f X ( x | B)
1
f X (x )
FX (x )

a x a x
(a) (b)
Fig. 4.4

Example 4.6: Let B represent the event { a < X (ξ ) ≤ b} with b > a.

For a given FX ( x ), determine FX ( x | B ) and f X ( x | B ).
Solution:
P { (X (ξ ) ≤ x ) ∩ (a < X (ξ ) ≤ b ) }
F X ( x | B ) = P { X (ξ ) ≤ x | B }=
P (a < X (ξ ) ≤ b )
P {(X (ξ ) ≤ x ) ∩ (a < X (ξ ) ≤ b )}
= . (22)
FX (b ) − FX (a )
For x < a, we have { X (ξ ) ≤ x}∩ { a < X (ξ ) ≤ b}= φ , and
hence FX ( x | B ) = 0. (23)
17
For a ≤ x < b, we have { X (ξ ) ≤ x }∩ {a < X (ξ ) ≤ b }= {a < X (ξ ) ≤ x }
and hence
P (a < X (ξ ) ≤ x ) F X ( x ) − F X ( a )
F (x | B) = = . (24)
FX (b ) − FX (a ) FX (b ) − FX (a )
X

For x ≥ b, we have { X (ξ ) ≤ x}∩ { a < X (ξ ) ≤ b}= { a < X (ξ ) ≤ b}

so that F X ( x | B ) = 1 . (25)
Using (23)-(25), we get (see Fig. 4.5)
 fX (x)
 , a < x ≤ b,
f X ( x | B ) =  FX (b ) − FX (a )

(26)
0, otherwise.

f X ( x | B)

f X (x )

x
a b
Fig. 4.5 18
We can use the conditional p.d.f together with the Bayes’
theorem to update our a-priori knowledge about the
probability of events in presence of new observations.
Ideally, any new information should be used to update our
knowledge. As we see in the next example, conditional p.d.f
together with Bayes’ theorem allow systematic updating. For
any two events A and B, Bayes’ theorem gives
P (B | A)P ( A)
P(A | B) = . (27)
P(B)

Let B = {x 1 < X ( ξ ) ≤ x 2 } so that (27) becomes (see (13) and (17))

P (( x 1 < X ( ξ ) ≤ x 2 ) | A )P ( A )
P {A | ( x 1 < X ( ξ ) ≤ x 2 ) }=
P (x 1 < X ( ξ ) ≤ x 2 )

∫
x2

F ( x 2 | A ) − F X ( x1 | A ) f X ( x | A ) dx
= X P ( A) = x1
P ( A ). (28)
F X ( x 2 ) − F X ( x1 )
∫
x2
f X ( x ) dx 19
x1
Further, let x1 = x, x2 = x + ε , ε > 0, so that in the limit as ε → 0,
lim P {A | ( x < X (ξ ) ≤ x + ε ) }= P (A | X (ξ ) = x ) =
f X ( x | A)
ε→0 fX (x)
P ( A ). (29)
or
P( A | X = x) fX (x)
f X |A ( x | A ) = . (30)
P ( A)

From (30), we also get

+∞ +∞
P ( A)∫ ∫

 
−∞
f X ( x | A ) dx =
1
−∞
P ( A | X = x ) f X ( x ) dx , (31)

or
+∞
P ( A) = ∫−∞
P ( A | X = x ) f X ( x ) dx (32)

and using this in (30), we get the desired result

P( A | X = x) fX ( x)
f X |A ( x | A ) = +∞
. (33)
∫
−∞
P ( A | X = x ) f X ( x ) dx 20
To illustrate the usefulness of this formulation, let us
reexamine the coin tossing problem.
Example 4.7: Let p = P(H ) represent the probability of
obtaining a head in a toss. For a given coin, a-priori p can
possess any value in the interval (0,1). In the absence of any
additional information, we may assume the a-priori p.d.f f P ( p)
to be a uniform distribution in that interval. Now suppose
we actually perform an experiment of tossing the coin n
times, and k heads are observed. This is new information.
How can we update f P ( p ) ?
Solution: Let A= “k heads in n specific tosses”. Since these
tosses result in a specific sequence, f ( p)
P

P ( A | P = p ) = p k q n−k , (34) p
0 1
21
Fig. 4.6
and using (32) we get
1 1 ( n − k )! k !
P ( A) = ∫ 0
P ( A | P = p ) f P ( p ) dp = ∫ 0
p k (1 − p ) n − k dp =
( n + 1 )!
. (35)

The a-posteriori p.d.f f P ( p | A) represents the updated

information given the event A, and from (30) f ( p | A)
P |A
P( A | P = p) fP ( p)
f P |A ( p | A ) =
P ( A)
p
( n + 1 )!
=
( n − k )! k !
p k q n−k , 0 < p < 1 ∼ β ( n , k ). (36) 0
Fig. 4.7
1

Notice that the a-posteriori p.d.f of p in (36) is not a

uniform distribution, but a beta distribution. We can use this
a-posteriori p.d.f to make further predictions, For example,
in the light of the above experiment, what can we say about
the probability of a head occurring in the next (n+1)th toss?
22
Let B= “head occurring in the (n+1)th toss, given that k
heads have occurred in n previous tosses”.
Clearly P ( B | P = p ) = p , and from (32)
1
P(B) = ∫ 0
P ( B | P = p ) f P ( p | A ) dp . (37)

Notice that unlike (32), we have used the a-posteriori p.d.f

in (37) to reflect our knowledge about the experiment
already performed. Using (36) in (37), we get
1 ( n + 1)! k +1
∫
k n−k
P(B) = p⋅ p q dp = . (38)
0 ( n − k )! k ! n+2
Thus, if n =10, and k = 6, then
7
P(B) = = 0 .58 ,
12
which is better than p = 0.5.
23
To summarize, if the probability of an event X is unknown,
one should make noncommittal judgement about its a-priori
probability density function f X (x ). Usually the uniform
distribution is a reasonable assumption in the absence of any
other information. Then experimental results (A) are
obtained, and out knowledge about X must be updated
reflecting this new information. Bayes’ rule helps to obtain
the a-posteriori p.d.f of X given A. From that point on, this
a-posteriori p.d.f f X |A ( x | A) should be used to make further
predictions and calculations.

24