Professional Documents
Culture Documents
6.1 Introduction
Exponential distributions have been introduced as a simple model for statistical analysis
of lifetimes. The bivariate exponential distribution and the multivariate extension of ex-
ponential distributions due to Marshall-Olkin (1967) has received considerable attention
in describing the statistical dependence of components in a 2-component system and in
developing statistical inference procedures. Moment generating function of the bivariate
generalized Exponential distribution is discussed by Ashour et al. (2009). Multivariate gen-
eralized Exponential distribution is studied by Mu and Wang (2010). Hanagal (1995) stud-
ied testing reliability in a bivariate exponential stress-strength model. A bivariate Marshall
and Olkin Exponential minification process is discussed by Ristic et al. (2008). Reliability
100
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
F (x)
G(x) = ; −∞ < x < ∞; 0 < α < ∞.
α + (1 − α)F (x)
αF (x)
G(x) = ; −∞ < x < ∞; 0 < α < ∞.
1 − (1 − α)F (x)
This new family involves an additional parameter α. In bivariate case if (X, Y ) be a random
vector with joint survival function F (x, y), then
F (x, y)
G(x, y) = ; −∞ < x < ∞; −∞ < y < ∞; 0 < α < ∞.
α + (1 − α)F (x, y)
constitute the Marshall-Olkin bivariate family of distributions. The new parameter α results
in added flexibility of distributions and influence the reliability properties.
Autoregressive models are developed with the idea that the present value of the se-
ries, Xt can be explained as a function of past values namely, Xt−1 , Xt−2 , ..., Xt−p where p
determines the number of steps in to the past, needed to forecast the current value. A first
order autoregressive time series model with exponential stationary marginal distribution
was developed by Gaver and Lewis (1980). Jose et al. (2011) introduced and studied a
Marshall-Olkin bivariate Weibull Process. In this chapter we are discussing three different
structures of the minification processes and develop minification process with Extended
Marshall-Olkin bivariate exponential distribution as marginal. First we consider a bivariate
101
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
where {(1n , 2n )} is a sequence of i.i.d. nonnegative random vectors, (X10 , X20 ) and
{(1i , 2i ), i ≥ 1} are independent random vectors, 0 < p < 1.
Now we consider a bivariate autoregressive minification process (Y1n , Y2n ) having the
structure {(Y1n , Y2n )} given by
(ε , ε ) w.p. α,
1n 2n
(Y1n , Y2n ) =
min(Y1n−1 , ε1n ), min(Y2n−1 , ε2n ), w.p. 1 − α
where 0 ≤ α ≤ 1.
Finally we consider a bivariate autoregressive minification process (Xn , Yn ) having
the structure {(Xn , Yn )} given by
(ε , η ) w.p. p,
n n
(Xn , Yn ) =
(min(Xn−1 , Yn−1 ), (εn , ηn )), w.p. 1 − p ,
where 0 ≤ p ≤ 1.
This chapter is arranged as follows. In section 6.2, Marshall- Olkin bivariate expo-
nential distribution and its properties are discussed. Extended marshall-Olkin bivariate
exponential distribution is introduced and studied in section 6.3. Multivariate extensions of
the models are given in section 6.4. Conclusions are given in section 6.5.
102
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Marshall and Olkin (1967) fatal shock model assumes that the components of a two-
component system die after receiving a shock which is always fatal. Independent Poisson
processes S1 (t, λ1 ) , S2 (t, λ2 ) , S3 (t, λ12 ) govern the occurrence of shocks. Events in the
process S1 (t, λ1 ) are shocks to component 1, events in the process S2 (t, λ1 ) are shocks
to component 2, and events in the process S3 (t, λ12 ) are shocks to both components. The
joint survival distribution (X1 , X2 ) of the components 1 and 2 is
103
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
where {(1n , 2n )} is a sequence of i.i.d. nonnegative random vectors. (ε1n , ε2n ) and
(X1m , X2m ) are independent random vectors for all m < n and 0 < p < 1.
Proof. First assume that the process {(X1n , X2n )} is a strictly stationary process with
MOBVE (λ1 , λ2 , λ12 ) distribution. Let F̄ (x1 , x2 ) be the survival function of the random
vector (X1n , X2n ) and Ḡ(x1 , x2 ) be the survival function of the random vector (ε1n , ε2n ).
Then it follows from model 1 that
This implies that the random vector (ε1n , ε2n ) has a MOBVE (λ1 , λ2 , λ12 ) distribution.
d
Conversely, assume that (ε1n , ε2n ) follows MOBVE (λ1 , λ2 , λ12 ) and (X10 , X20 ) =
(ε11 , ε21 ). Let F̄n (x1 , x2 ) be the survival function of the random vector (X1n , X2n ). Then
for n = 1 we have
104
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
which implies that (X11 , X21 ) has a MOBVE (λ1 , λ2 , λ12 ) distribution. Suppose now that
(X1i , X2i ) has a MOBVE (λ1 , λ2 , λ12 ) distribution, i = 1, 2, . . . , n − 1. Then
d d
i.e. (X1n , X2n ) = MOBVE(λ1 , λ2 , λ12 ). Thus (X1n , X2n ) = (X10 , X20 ) for every n and
since the process {(X1n , X2n )} is a Markov process, it follows that the process {(X1n , X2n )}
is a strictly stationary process.
First we shall consider the joint survival function of the random vectors (X1n+h , X2n+h )
and (X1n , X2n ). From model 1 the joint survival function is given by
S̄h (x1 , x2 , z, v) = P (X1n+h > x1 , X2n+h > x2 , X1n > z, X2n > v)
= S̄h−1 (px1 , px2 , z, v) Ḡ ((1 − p)x1 , (1 − p)x2 )
h
h h
Y
Ḡ ph−i (1 − p)x1 , ph−i (1 − p)x2
= S̄0 p x1 , p x2 , z, v
i=1
h
F̄ (p(h−i) (1 − p)x1 , p(h−i)/ (1 − p)x2 )
Y
= S̄0 ph x1 , ph x2 , z, v
i=1
F̄ (p(h−i+1) (1 − p)x1 , p(h−i+1) (1 − p)x2 )
F̄ max(ph x1 , z), max(ph x2 , v) F̄ ((1 − p)x1 , (1 − p)x2 )
=
F̄ (ph (1 − p)x1 , ph (1 − p)h x2 )
We can see that the joint survival function has an absolutely continuous component for
z > ph x1 , v > ph x2 , z 6= v and x1 6= x2 .
We now consider the autocovariance structure of the Marshall–Olkin bivariate Expo-
nential minification process. Let us first consider the autocovariance function of the random
105
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
P X1n+1 = p−1 X1n |X1n = x1 = P ε1n+1 ≥ p−1 (1 − p)x1 = e−(λ1 +λ12 )(1−p)x/p .
(6.2.4)
Now, using equations (6.2.3) and (6.2.4) the conditional expectation is obtained as,
p2
E(X1n+1 |X1n = x1 ) =
(λ1 + λ12 )2 (1 − p)
Using the fact that E(X1n+1 X1n ) = E(X1n E(X1n+1 |X1n )), we get
1 − p2
E(X1n+1 X1n ) =
(λ1 + λ12 )2 (1 − p)
p
Cov(X1n+1 , X1n ) =
(λ1 + λ12 )2
p
Similarly, we have Cov(X2n+1 , X2n ) = (λ2 +λ12 )2
. Let us consider now the autocovari-
ance function of random variables X1n+1 and X2n . We need the joint survival function
P (X1n+1 > x1 , X2n > x2 ). It can be derived as
P (X1n+1 > x1 , X2n > x2 ) = P (X1n > px1 , X2n > x2 ) P (ε1n+1 > (1 − p)x1 )
= e−(λ1 +λ12 (1−p))x1 −λ2 x2 −λ12 max(px1 ,x2 ) .
106
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Then similar to the derivation of Cov(X1n+1 , X1n ) we can obtain Cov(X2n+1 , X2n ).
The autocovariance matrix at lag 1 of a Marshall–Olkin bivariate exponential minifica-
tion process is
p pλ12
(λ1 +λ12 )2 (λ1 +λ12 )(λ2 +λ12 )(λ1 +λ12 +λ2 p)
C= pλ12 p
.
(λ1 +λ12 )(λ2 +λ12 )(λ2 +λ12 +λ1 p) (λ2 +λ12 )2
we can take the estimates of the parameters λ1 , λ2 and λ12 as the solutions of the system
of the equations
n
!−1
1 X
λ
b1 + λ
b12 = Xi
n + 1 i=0
n
!−1
1 X
λ
b2 + λ
b12 = Yi
n + 1 i=0
n
1 1 1 1 X
+ = Xi Yi .
λ
b1 + λ
b2 + λ
b12 λ b1 + λ
b12 λ b2 + λ
b12 n + 1 i=0
107
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Figure 6.1: The simulated sample path for various values of n and p when λ1 = 0.5,
λ2 = 1, λ12 = 1.5
In order to study the behavior of the processes, we simulate the sample paths for various
values of n and p. In particular we take λ1 = 0.5, λ2 = 1 and λ12 = 1.5 and is given in
Figure 6.1. In Fig. 1a, Fig. 1b and in Fig. 1c we take n = 200 and p = 0.6. In Fig. 2a, Fig.
2b and in Fig. 2c we take n = 300 and p = 0.8. In Fig. 3a, Fig. 3b and in Fig. 3c we take
n = 400 and p = 0.9.
Mukherjee and Maiti (2005) discussed the determination of reliability with respect to MOBVE
distribution. They consider the survival function given in (6.2.1) and also consider Y
(strength) as non negative random variable following the exponential distribution with sur-
vival function
Ḡ(y) = e−y ; 0 < y < ∞
108
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
λ1 + λ2 + λ12
R=
1 + λ1 + λ2 + λ12
1 1 1
R = [1 − − + ]
λ1 + λ12 + 1 λ2 + λ12 + 1 λ1 + λ2 + λ12 + 1
Theorem 6.3.1. Let {(X1n , X2n ), n ≥ 1} be a sequence of i.i.d. random vectors with
common survival function F̄ (x1 , x2 ) which is survival function of MOBVE(λ1 , λ2 , λ12 ).
Let N be a random variable with a geometric(α) distribution and suppose that N and
(X1i , X2i ) are independent for all i ≥ 1. Define UN = min X1i and VN = min X2i .
1≤i≤N 1≤i≤N
Then the random vector (UN , VN ) is distributed as EMOBVE (α, λ1 , λ2 , λ12 ) if and only
if (X1i , X2i ) has MOBVE(λ1 , λ2 , λ12 ) distribution.
∞
X
S̄(x1 , x2 ) = P (UN > x1 , VN > x2 ) = [F̄ (x1 , x2 )]n (1 − α)n−1 α
n=1
αF̄ (x1 , x2 ) αe−λ1 x1 −λ2 x2 −λ12 max(x1 ,x2 )
= =
1 − (1 − α)F̄ (x1 , x2 ) 1 − (1 − α)e−λ1 x1 −λ2 x2 −λ12 max(x1 ,x2 )
109
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Model 1
Consider a bivariate minification process {(X1n , X2n ), n ≥ 0} given by
where {(1n , 2n )} is a sequence of i.i.d. nonnegative random vectors, (X10 , X20 ) and
{(1i , 2i ), i ≥ 1} are independent random vectors, 0 < p < 1.
Proof. Let F̄ (x1 , x2 ) be the survival function of the random vector (X1n , X2n ) and Ḡ(x1 , x2 )
be the survival function of the random vector (1n , 2n )
This implies that {(1n , 2n )} has EMOBVE (λ1 , λ2 , λ12 ) distribution.
Conversely assume that {(1n , 2n )} has EMOBVE (λ1 , λ2 , λ12 ) distribution.
Then
F̄ (x1 , x2 ) = e(−λ1 x1 −λ2 x2 −λ12 max(x1 ,x2 ))
Model 2
Consider a bivariate autoregressive minification process (Y1n , Y2n ) having the structure
110
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
where 0 ≤ α ≤ 1.
Theorem 6.3.3. A minification process {(Y1n , Y2n ), n ≥ 0} has EM OBV E(α, λ1 , λ2 , λ12 )
stationary marginal distribution if and only if (1n , 2n ) has M OBV E(λ1 , λ2 , λ12 ) dis-
tribution and (X0 , Y0 ) has MOBVE (α, λ1 , λ2 , λ12 ) distribution .
Proof. Let Ḡn (x1 , x2 ) and F̄ (x1 , x2 ) be the survival functions of (X1n , X2n ) and (1n , 2n ),
respectively. From the definition of the process, we have that
Ḡn (x1 , x2 ) = P (Y1n > x1 , Y2n > x2 ) = αF̄ (x1 , x2 ) + (1 − α)Ḡn−1 (x1 , x2 )F̄ (x1 , x2 ).
(6.3.1)
Under stationarity
Replacing Ḡ with the survival function of the random vector with EMOBVE (α, λ1 , λ2 , λ12 )
distribution and solving the obtained equation on F̄ (x1 , x2 ), we obtain
which is the survival function of EMOBVE (α, λ1 , λ2 , λ12 ). Hence it follows that (Y11 , Y21 )
d
has EMOBVE(α, λ1 , λ2 , λ12 ). Now assume that (Y1n−1 , Y2n−1 ) = EMOBVE(α, λ1 , λ2 , λ12 ).
111
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Then
Thus, (Y1n , Y2n ) follows EMOBVE(α, λ1 , λ2 , λ12 ) distribution. Hence by induction (Y1n , Y2n )
has EMOBVE(α, λ1 , λ2 , λ12 ) distribution for every n≥ 0. This establishes stationarity.
Corollary 6.3.1. If (Y10 , Y20 ) has an arbitrary bivariate distribution and {(1n , 2n )}
has MOBVE (λ1 , λ2 , λ12 ) distribution, then {(Y1n , Y2n )} has EMOBVE (α, λ1 , λ2 , λ12 )
distribution asymptotically.
n−1
X
= αF̄ (x1 , x2 ) (1 − α)j F̄ j (x1 , x2 ) + (1 − α)n F̄ n (x1 , x2 )Ḡ0 (x1 , x2 )
j=0
αF̄ (x1 , x2 ) 1 − (1 − α)n F̄ n (x1 , x2 )
= + (1 − α)n F̄ n (x1 , x2 )Ḡ0 (x1 , x2 ).
1 − (1 − α)F̄ (x1 , x2 )
Model 3
Consider a bivariate autoregressive minification process (Xn , Yn ) having the structure
112
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
{(Xn , Yn )} given by
(ε , η ) w.p. p,
n n
(Xn , Yn ) =
(min(Xn−1 , Yn−1 ), (εn , ηn )), w.p. 1 − p ,
Proof. Let Ḡ(x, y) and F̄ (x, y) be the survival functions of (Xn , Yn ) and (n , ηn ) respec-
tively. Under stationarity
if we take
F̄ (x, y) = e−λ1 x−λ2 y−λ12 max(x,y) ,
we get
pe−λ1 x−λ2 y−λ12 max(x,y)
Ḡ(x, y) =
1 − (1 − p)e−λ1 x−λ2 y−λ12 max(x,y)
Conversely,
Ḡ(x, y)
F̄ (x, y) = ,
p + (1 − p)Ḡ(x, y)
then we have
F̄ (x, y) = e−λ1 x−λ2 y−λ12 max(x,y) ,
Let X1 and X2 be two non negative random variables jointly following EMOBVE distribution
with survival function
113
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Also let Y( Strength) be a non-negative random variable following the exponential distribu-
tion with Survival function
G(y) = e−y ; 0 < y < ∞
Our objective is to derive stress-strength reliability R when stress (X) has two components
X1 and X2 and strength Y is independently distributed.
Case(i): Stress components are in series
We define,
U = min(X1 , X2 ).
R = P (U < Y )
Z ∞ Z y
= { f (u)du}g(y)dy
0 o
Z ∞ −y
e − e−(1+λ1 +λ2 +λ12 )y
= dy
0 1 − (1 − α)e−(λ1 +λ2 +λ12 )y
From the Table 6.1, it is clear that as α increases reliability decreases. Using this systems
with optimal reliability values can be designed.
Case(ii): Stress components are in parallel
In this case, we consider
V = max(X1 , X2 )
114
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
α R
0.005 0.995763
0.500 0.898504
1 0.857142
2 0.807543
3 0.775411
5 0.732582
10 0.672124
50 0.534067
100 0.479660
2000 0.200940
Then,
R = P (V < Y )
Z ∞ Z y
= { f (v)dv}g(y)dy
0 o
Z ∞
= FV (y)e−y dy
0
From the Table 6.2, it is clear that as α increases reliability decreases. Using this
systems with optimal reliability values can be designed.
115
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
α R
0.05 0.960070
0.50 0.838880
2 0.703029
3 0.656812
5 0.596680
20 0.436225
100 0.280610
1000 0.133670
failure time of the ith component. The joint distribution of lifetimes (t1 , ..., tk ) is given by the
Marshall-Olkin multivariate exponential distribution. The joint survival function of lifetimes
is given by
k
X k
X
P (t1 > x1 , ..., tk > xk ) = exp(− λi x i − λij max{xi , xj }−λ1,...,k max{x1 , ..., xk })
i=1 i,j=1
Multivariate Extended Marshall-Olkin bivariate exponential distribution can be written
k k
!
X X
α exp − λi x i − λij max{xi , xj } − λ1,...,k max{x1 , ..., xk }
i=1 i,j=1
Ḡ(x1 , ..., xk ) = k k
!
X X
1 − (1 − α) exp − λi x i − λij max{xi , xj } − λ1,...,k max{x1 , ..., xk }
i=1 i,j=1
116
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Let
(ε , ...ε ) w.p. α,
1n pn
Y n = {Y1n ...Ypn } =
min(Y1n−1 , ε1n ), ... min(Ypn−1 , εpn ), w.p. 1 − α
where 0 < α < 1. Then Y n have extended multivariate Marshall Olkin exponential
stationary marginal distribution if and only if (ε1n , ...εpn ) has multivariate Marshall Olkin
exponential distribution.
6.5 Conclusions
Extended Marshall-Olkin bivariate exponential distribution is introduced and its properties
are studied. Expressions for stress-strength reliability of a two component system are
derived. Reliability R for various parameter combinations are also computed. From the
tables it is clear that as α increases reliability decreases. Using this systems with optimal
reliability values can be designed. Multivariate extensions are also done to model multi-
component systems. We introduced three different forms of minification processes and
necessary and sufficient conditions for stationarity are established.
References
Ashour, S.K., Amin, E.A., Muhammed, H.Z. (2009) Moment Generating Function of the
Bivariate Generalized Exponential distribution, Applied Mathematical Sciences, 3,
59 (2911-2918).
Gaver, D.P., Lewis, P.A.W. (1980) First order autoregressive gamma sequences and point
processes, Adv. Appl. Pro., 12, 727-745.
Jose, K.K., Ancy Joseph, Ristic, M.M. (2011) Marshall Olkin Weibull Distributions and
Minification Processes, Statistical Papers, 52, 789-798.
117
C HAPTER 6. M ARSHALL -O LKIN B IVARIATE E XPONENTIAL D ISTRIBUTION : G ENERALISATIONS AND A PPLICATIONS
Marshall, A. W., Olkin, I. (1997) A new method for adding a parameter to a family of dis-
tributions with application to the exponential and weibull families, Biometrica 84(3),
641-652.
Mokhlis, N.M. (2006) Reliability of strength model with a bivariate exponential distribution,
Journal of the Egyptian Mathematical Society, 14(1), 69-78.
Mukherjee, S.P., Maiti, S.S. (2005) Stress -strength Reliability under two component
stress system, Journal of Statistical Theory and Applications, 341-347.
Ristic, M.M., Popovic, B.C., Nastic, A., Dordevic, M. (2008) Bivariate Marshall Olkin Ex-
ponential Minification Processes, Filomat 22, 1, 67-75.
118