International Research Journal of Applied and Basic Sciences

© 2013 Available online at www.irjabs.com

ISSN 2251-838X / Vol, 4 (6):1530-1534
Science Explorer Publications

A new approach on the renewal process with

geometric interarrival times
H. Samimi and Mohammad Mohammadi
Department of Statistics, Faculty of Mathematical Sciences, University of Guilan, Rasht, Iran.

Corresponding author email: [email protected]

ABSTRACT: This paper develops the probability of a renewal process, whose interarrival are
independent and identically distributed (i.i.d.) random variables (R.V.s) with geometric distribution.
The distribution of the number of renewals in (0, t ] and the joint distribution of the number
renewals in (0, a] and [a, b] , 0 < a < b , are obtained. Then, using the joint distribution, the distribution
of the number of renewals in [a, b] , 0 < a < b , is obtained. Finally , we show that every renewal
processes, whose interarrivals are i.i.d. discrete R.V.s, has independent and stationary increment if
and only if its interarrivals be geometric distribution.
Keywords: Renewal process, Independent increment, Stationary increment, Geometric distribution.

1. INTRODUCTION
A renewal process {N t : t ≥ 0} is a nonnegative integer valued stochastic process that registers the
successive occurrence of an event during the time interval (0, t ] , where the time duration between consecutive
n
"event" are positive i.i.d. R.V.s. Let the successive interarrival times be T 1 ,T 2 , … and Sn = T i for n ≥ 1 ,
i =1
S 0 = 0 , (Karlin and Taylor, 1975). Suppose A is a bounded Borel subset of (0, ∞). N A denotes the number
of the elements of the set {n > 0 : S ∈ A }.
n
In the case A = ( 0, t ] N A by N t . It is well known that if interarrival times in a renewal
we denote
process are i.i.d. exponential R.V.s, the N A has poisson distribution with parameter λ m ( A ) , m denoting the
Lebesgue measure. In particular for A = [a, b] , 0 < a < b , N A has poisson distribution with parameter
λ (b − a ) , (Kingman, 1993 ,. Varsei , 2008) . In literature different distributions are considered as interarrival
times in renewal processes. Cuffe and Friedman (2006) obtained the exact distribution of N t , when interarrival
times are sums of two independent exponential R.V.s with likely unequal parameters. Rinne ( 2009) studied
renewal process with Weibul interarrival times. Parzen (1999) and Varsei and Samimi ( 2009) applied the Erlang
distribution as interarrival times of renewal process, However in almost all of them, the distribution of N t and its
expectation was argued but the distribution of N A , A = [a, b ] , 0 < a < b , was not obtain. The distribution of
N A in the case Erlang interarrival time was obtained (Parzen, 1999).
N t and E ( N t ) , and then we obtain distribution of
In this paper, in section 2, we obtain the distribution
N A , A = [a, b] , 0 < a < b , by using the joint distribution of ( N a , N A ) in section 3. Then section 4, we
discusses product measure of discrete random vector and necessary and enough condition for stationary and
independent increment.
Intl. Res. J. Appl. Basic. Sci. Vol., 4 (6), 1530-1534, 2013

2..>, . DISTRIBUTION OF Nt
We assume that the interarrival times T1 ,T 2 , … in a renewal process are discrete R.V.s,
having a geometric distribution with probability mass functi on (p.m.f.) P (T i = x ) = pq x −1 ,
i = 1, 2,…, x = 1, 2 ,…, p + q = 1, 0 < p < 1 and cumulative distribution function F (x ) = 1 − q x.

Lemma
n
Suppose S n = T i for n ≥ 1 , S 0 = 0 and F ( x ) = P (S n ≤ x ), then
i =1
n −1
x p
Fn ( x ) = 1 − q x ( q )i , x = n , n + 1,… (1)
i =0 i
Proof
We shall the result using mathematical induction and the n-fold convolution of F ( x ), for discrete R.V.s
x
is given by Fn ( x ) = Fn −1 ( x − u ) P ( X = x ).
u =0
It can be obtained the p.m.f. of N t for t ∈ , by (1), we have

P ( N t = n ) = P ( N t ≥ n ) − P ( N t ≥ n + 1) = Fn (t ) − Fn +1 (t )
t (2)
= p n q t −n , n = 0,1, …, t
n
It flows Nt, t ∈ has binomial distribution with parameters (t , p ) .

3. . JOINT jDISTRIBUTION OF  Na , NA 
In the renewal process let T1 ,T 2 , … be i.i.d. geometric R.V.s with parameter p . Let A = [a, b] ,
0 < a < b . To obtain the distribution of N A , we first calculate the joint distribution ( N a , N A ) .

Theorem1
In the renewal process the joint distribution of ( N a , NA ) is equal to
a b − a b p k +s
P (N a = k , N A = s ) = q ( ) , k , s = 0,1,…, k + s ≤ b (3)
k s q
Proof

For different value of k and s have

P ( N a = 0, N A = 0 ) = P ( N b = 0 ) (4)

P ( N a = 0, N A = 1) = P ( a < S 1 < b < S 2 ) (5)

P ( N a = 0, N A = s ) = P ( a < S 1 < S s < b < S s +1 ) ,s ≥ 2 (6)
P ( N a = k , N A = 0 ) = P ( S k < a < b < S k +1 ) ,k ≥1 (7)

P ( N a = k , N A = 1) = P ( S k < a < S k +1 < b < S k + 2 ) ,k ≥1 (8)

P ( N a = k , N A = s ) = P ( S k < a < S k +1 < S k +s < b < S k +s +1 ) , k ≥ 1, s ≥ 2 (9)

The equality (4) is obtained by equation (2). Hence in the case k = s = 0 , the theorem holds, now we
prove the theorem in the case k ≥ 1 and s ≥ 2 , other case are obtained similarly.

1531
Intl. Res. J. Appl. Basic. Sci. Vol., 4 (6), 1530-1534, 2013

Let
X 1 = S k =Y 1
X 2 = S k +1 =Y 1 +Y 2
X 3 = S k +s =Y 1 +Y 2 +Y 3
X 4 = S k +s +1 =Y 1 +Y 2 +Y 3 +Y 4
where
Y 1 ~ NB ( k , p ) , Y 2 ~ Geo ( p ) , Y 3 ~ NB ( s − 1, p ) and Y 4 ~ Geo ( p ) and Y 1 , Y 2 , Y 3 and Y 4 are
independent.

The value of equation (9) equals

x 3 +1
∞ b a
x1 x 3 − x 2 − 1 k +s +1 x 4 − k −s −1
p q (10)
x 4 =b +1 x 3 =a + 2 x 2 =a +1 x 1 = k k −1 s −2
We can calculate the summation (10) and obtain equation (3) for k ≥ 1 and s ≥ 2 , by using the relations
a
n +i n + a +1 b x 3 −1
x 3 − x 2 −1 b −a
= and = .
i =1 n a x 3 =a + 2 x 2 =a +1 s −2 s

Corollary 1
The distribution of N A in the renewal process is
a
a b −a p b − a s (b −a )−s
P (N A = s ) = ( ) k +s q b = pq (11)
k =0 k s q s
The expectation of N A is obtained by (11), we have E ( N A ) = (b − a ) p .

4 . CHARACTERIZATION OF THE RENEWAL PROCESS

At first, we review some concepts about product measure. Let T and W
are discrete random vectors on measure spaces (X ,  ,  ) , (Y ,  , ) , then (T,W ) has distribution
π = µ ×ϑ in χ × γ ( µ × ϑ is product measure and χ × γ is σ -field that generated by the measureable

rectangles). We have π ( E ) = π ' ( E ) = π "( E ) , E ∈ χ × γ , where

π '(E ) = ϑ y : ( x , y ) ∈ E µ {x } ,
{ x : x ∈X }

π "( E ) = µ x : ( x , y ) ∈ E ϑ { y }.
{ y : y ∈Y }

Remark1
Suppose B ∈ χ × γ and A ∈ χ then
π ( ( A ×Y ) B)= ϑ y : ( x , y ) ∈ B µ{x } (12)
{x : x ∈A }
Remark2
If T1 and T2 are independent random vectors with distribution µ and ϑ in χ and γ respectively, then
P ((T1 ,T 2 ) ∈ B ) = P ( (t 1 ,T 2 ) ∈ B ) µ {x } ,B ∈ χ ×γ (13)
{x : x ∈X }

P (T1 ∈ A , (T1 ,T 2 ) ∈ B )) = P ( (t 1 ,T 2 ) ∈ B ) µ {x } ,B ∈ χ ×γ , A ∈ χ (14)

{x : x ∈A }
Now, let N t is a renewal process characterized by the fact that successive interarrival times T1 ,T 2 , … are
i.i.d. nonnegative integer R.V.s.

1532
Intl. Res. J. Appl. Basic. Sci. Vol., 4 (6), 1530-1534, 2013

Condition (i ): For each ω , N t (ω ) is a nonnegative integer for t ≥ 0, N0 (ω ) = 0 and lim N t ( ) = ∞ ,

t →∞
further, for each ω , N t (ω ) as a function of t is non-decreasing and right-continuous and the point of discontinuity
the N t ( ω ) − sup s < t N s (ω ) is exactly 1.

Condition (ii ): The T 1 ,T 2 , … are independent, having a geometric distribution with parameter p.
Condition (iii ): (a) For 0 < t 1 < t 2 < … < t k increment N t1 , N t 2 − N t1 ,… , N t k − N t k −1 are independent.
(b) The increment s have Negative binomial i.e.

t −s
P (N t − N s = n ) = p n (1 − p )t −s −n , 0 ≤ s < t , n = 0,1,… (15)
n

Theorem 2

Condition (ii ) and (iii ) are equivalent in the presence of condition (i ).

Proof of (ii ) (iii )

Fix t, and consider the events that happen after time t. By N t = max {n : S n ≤ t } and
S N t ≤ t < S N t +1 ,

we have
T1(t ) = S N t +1 − t , T 2(t ) = T N t + 2 , T 3(t ) = T N t +3 , …
T 1(t ) ,T 2 (t ) ,... define the waiting times following t. By [ N t ≥ n ] = [S n ≤ t ] , or N t +s − N t ≥ m , or
N t +s ≥ N t + m if and only if S N t + m ≤ t + s , which is the same thing as T 1(t ) + … + T m (t ) ≤ s Thus
N t +s − N t = max{m :T 1(t ) + …+ T m (t ) ≤ s } (16)
Hence
[ N t +s − N t = m ] = { T 1(t ) + … + T m (t ) ≤ s < T 1(t ) + …+ T m +1(t ) } (17)
Suppose µn is measure function of S n and y ≥ 0 , by using equation (14), we have
P ( S n ≤ t < S n +1 , S n +1 − t > y ) = P ( S n ≤ t , T n +1 > t + y − S n )
= P (T n +1 > t + y − x ) µ n {x } = q y P (T n +1 > t − x ) µ n {x }
{x : x ≤ t } {x : x ≤ t }

= q y P (S n ≤ t , T n +1 > t − S n )
By using independent of the T 1 ,T 2 , …
P ( S n +1 − t > y 1 , T n + 2 > y 2 ,… , T n + j > y j , S n ≤ t < S n +1 )
yj
= P ( S n +1 − t > y 1 , S n ≤ t < S n +1 ) q y 2 …q
yj
= P ( S n ≤ t < S n +1 ) q y 1q y 2 …q

If H = ( y 1 , ∞ ) × …× ( y j , ∞ )
Thus
P ( N t = n , (T 1(t ) ,…, T j (t ) ) ∈ H ) = P ( N t = n ) P ((T 1 , …, T j ) ∈ H ) (18)

Now the event [ N s i = m i , 1 ≤ i < u ] can be put in the form [(T 1 , … ,T j ) ∈ H ] , where j = mu + 1 and H
j
is the set of x in for which x 1 +…+ x m i ≤ s i < x 1 + …+ x m i +1 , 1 ≤ i ≤ u Thus equation (18) gives

( )
P N t = n , N t +s i − N t = m i , 1 ≤ i ≤ u = P ( N t = n ) P (N s i = m i , 1 ≤ i ≤ u )

1533
Intl. Res. J. Appl. Basic. Sci. Vol., 4 (6), 1530-1534, 2013

From this it follow by induction on k that if 0 = t 0 < t 1 < … < t k , then

k

( )
P N t i − N t i −1 = n i , 1 ≤ i ≤ k = ∏ P ( N t i −t i −1 = n i ) (19)
i =1
Thus condition (ii ) implies equation (19). But from equation (19) and equation (2) follow the two parts of
condition (iii ).

Proof of (iii) (ii)

If (iii ) holds, then P (T1 > t ) = q t , so that T1 has geometric distribution. To find the joint distribution
of T1 and T2 , suppose that 0 ≤ s1 < t 1 < s 2 < t 2 and perform the calculate
P ( s1 < S 1 ≤ t1 , s 2 < S 2 ≤ t 2 )
= P (N s1 = 0, N t1 − N s1 = 1, N s 2 − N t1 = 0, N t 2 − N s 2 ≥ 1)
= q s1 (t 1 − s 1 ) pq ( 1
t − s1 )
q(
s 2 −t 2 )
(1 − q ( t 2 −s 2 )
) = p (t 1 − s1 ) (q s 2 −1 − q t 2 −1 )
= p 2q y 2 − 2
{( y 1 , y 2 ): s1 < y 1 ≤t1 , s 2 < y 2 ≤t 2 }

Thus for a rectangle A contained in the open set G = [( y 1 , y 2 ) : 0 < y 1 < y 2 ]

µ ' ( A ) = P ( (S 1 , S 2 ) ∈ A ) = p 2q y 2 − 2
{( y 1 , y 2 ):( y 1 , y 2 )∈A }

( ' is measure function of ( S 1 , S 2 ) )

By inclusion – exclusion, this holds for finite unions of such rectangle and hence, by using continuity from
below, for countable ones. Therefore, it holds for A =G
G ' if G ' is open. Since the open sets form as
- system generating the Borel sets, ( S 1 , S 2 ) has mass p q 2 on G . By a similar argument in
2 y −2 k
,
(S 1 ,…, S k ) has mass p q on [ y : 0 < y 1 < … < y k ] . If a linear transformation g ( y ) = x is defined
k y k −2

k
by x i = y i − y i −1 then (T1 ,…,T k ) = g (S 1 , …, S k ) has mass ∏( pq y i −1
) . This proves condition (ii ).
i =1

CONCLUSIONS

In this paper we consider the renewal process whose interarrival times are i.i.d. geometric distribution. We
show that the renewal processes has independent and stationary increments, NA has binomial distribution by
a new method. Also, we show if, a renewal process has with discrete integer interarrivals has independent and
stationary increments. Then its interarrivals time has geometric distribution.

REFERENCE

Karlin S, Taylor HM. 1975. A First Course in Stochastic Processes. Academic Press.
Kingman JFC. 1993. Poisson Processes. Oxford studies in Probability.
Varsei A . 2008. On the Poisson process in 2. Far East Journal of Theoretical Statistics . 25 (1): 95-101.
Cuffe BP, Friedman MF. 2006. On the exact distribution of delayed renewal process with exponential sum interval times. European Journal
of operational research. 172: 909-918.
Rinne H. 2009. The Weibull Distribution. Taylor & Francis Group, LLC.
Varsei A, Samimi H. 2009. Some results on renewal process with Erlang interarrival times. Applications and Applied Mathematics: An
International Journal. 4 (1): 88-97.
Parzen E . 1999. Stochastic processes. SIAM's classics in Applied Math.

1534