24 views

Uploaded by Arshiya Jabeen

© All Rights Reserved

- Discrete Probability Distribution
- Bayesian Methods for Hackers
- STAT 3502 Outline Winter 2015
- Business Analytics by James Evans TOC
- December 09 10
- Screen ns Lab 5
- October 2010
- Theory of Probability
- MATH1151 Course Outline
- 00 789814730761.pdf
- 15MA301-U2
- Queing Theory Model.ppt
- ch3Harrell Ghosh Bowden cap2
- Poisson Process
- Risk Avoidance in Insurance
- Final
- 13a S1 January 2014
- ps19ans
- CHAPTER 5 Reviewer (1) (1) (1)
- l3 probability methods

You are on page 1of 27

2.1 Introduction

utilised for this thesis. We start with the well known lack of memory

property and then its multivariate extensions. While carrying out an

extensive study, it is seen that, there are several situations where the

LMP is not satisfied but properties which are very near to LMP is

satisfied. This was one of the motivations for extending the LMP. In a

sense the SCBZ property is an extension of the LMP.

its future performance does not depend on the past, knowing its present

condition ,

LMP. In proper

14

Lebesgue measure, we say that the random variable X or its

distribution has LMP if for all x, y?O,

P(X>_x+yJ X>_y)=P(X>_x) (2.1)

with P(X>_ y) strictly greater than zero or equivalently.

P(X_ x+ y)=P( X? X) P(X_y) (2.2)

for all x, y>_0 and P(X=0)#1.

due to its application to real life situation. If R(x)=P(X_x) denote the

survival function of the random variable X. then in terms of R(), the

LMP can be stated as

R(x+y) = R(x) R(y) (2.3)

Galambos and Kotz (1978) have established the equivalence of

LMP, constancy of failure rate and constancy of mean residual life.

Cauchy (1821) and Darboux (1875) have established that the unique

non-zero solution of (2.3) is R(x) = e-"" for some constant X, which is

the survival function of the exponential distribution. So exponential

distribution is the only distribution having this property.

extensions in various directions suggested by different authors. Some

of them are listed here.

Lebesgue measure, then Fortet (1977 ) had shown that this amounts to a

characterization of the exponential distribution .

Sethuraman (1965)

finite induction as given below . The equation (2.3) can be written as

R(xl+x2+... +x,) = R(xi) ...R(xr,)

(2.4)

R(nx)=[ R(x)]n

If (2.4) holds for any two integers nl and n2 such that

(2.5)

loge, is

1og n2

and Tubilla (1975) showed that (2.3) is valid for two values y, and y2

of y such that 0< y'< y2 and y,/y2 is irrational for all x>0, then X

follows a negative exponential.

form of the conditional expectation of a function specifically,

E(h(X)JX>

_x) = g(x), x>_0 (2.6)

or

E(h(X-x)IX>_x) = g(x),

x?O (2.7)

Here h(.) and g (.) are known functions ending up with the

solutions that are proper survival functions. More details are available

in Kotlarski (1972), Laurent (1972 ), Shanbhag and Rao (1975), Dallas

( 1976) and Gupta (1976).

manner . In place of (2.2) they used the equation

P(X> x* Y)= P(X> X) P( X> Y), (2.8)

being taken as an associative and reducible binary operator. In this

case (2.3) reduce the form

R(x*y) = R(x)R(y).

(2.9)

x*Y = g 1(g(x)+g(y))

(2.10)

Dimitrov (1992) introduced the concept of almost lack of memory

property (ALMP). X is said to have the ALMP if there exists a

P(X>_b+xjX>_b) = P(X>_x)

holds for any b = a,,, n=1,2, ... and for all x>_0.

Galambos and Kotz (1978) bring out the equation

(2.11)

to characterize the uniform distribution in [0,1] of X. Also the equation

P(X>_uvIX>_v)= P(X>_u), u, v>1 (2.13)

characterizes the Pareto distribution with survival function

R(x) = x-', x>_1, r>O. (2.14)

(2.12) and (2.13) are called the multiplicative lack of memory

property. Dimitrov and Collani (1995) introduced the multiplicative

almost lack of memory property.

lack of memory property of type I (MALMI) if there exists a sequence

of numbers {v.}n, , 0<_v<_1, V:#Vm for nom such that

P(X<_uvIX<_v) = P(X5u) (2.15)

for all u>_0. A random variable X is said to have multiplicative almost

lack of memory property of type 2 (MALM2) if there exists a sequence

{vn n=1

} , v?1, v>_ vm for all n#m such that

P(X>_uvIX>_V) = P(X>_u)

(2.16)

A somewhat different approach to the extension was considered

by Huang ( 1981 ). If X and Y are two independent non-negative random

variables, we say that X is ageless relative to Y if P(X>Y)>O and

10

for all x>_O.

integer valued random variable satisfying the condition

P(X>_x+yJ X>_y)=P(X>_x)

then X follows a geometric distribution with

P(X=x) = p(1-p)X, x =0,1,2,...; 0<p<l. (2.18)

The geometric random variable is the only discrete random variable

having the LMP.

failure rate and constancy of mean residual life function. The

characterization of geometric distribution and discrete IFR (DFR)

using order statistics are established in Neweichi and Govindarajulu

(1979). Another characterizations of geometric distribution in terms of

order statistics are studied in Arnold (1980) and Srivastava (1974).

Nair and Hitha (1989) characterizes the discrete models using

distributions on partial sums. The motivation behind this was the work

of Xekalaki (1983).

11

extended to higher dimensions . Let us first present some important

variations in the bivariate cases.

{(xl,x2 ) :

R(xl+yl, x2+y2) =

(2.19)

where

R(xl,x2) =

P(Xl_

xl,X2?x2)

R(xi,x2) = e- x1-A2x2 (2.20)

which is the product of the marginals. If (X1iX2) is taken to be the

lifetimes of a two component system, (2.20) shows that the life times

of the components are independent, which does not have any relevance

in life testing. To introduce the-dependency of the component life

times we consider the equation

R(xl+t,x2+t) = R(t,t) R(xi,x2) (2.21)

for all x1,x2, t>_0.

12

as

R(xj,x2) = exp(-21x1-22x2-212 max(xl ,x2)). (2.22)

Here also the marginal distributions are again exponential but they are

not independent . Also P(Xl=X2)>O.

continuous .

distribution with independent exponential marginals .

absolute continuity in which marginals are not purely exponential but

are mixtures of exponentials.

investigate the behaviour of one of the components, when lifetime of

the other is pre-assigned. The first work in this direction is due to

Johnson and Kotz (1975) who defined the vector valued failure

rate(hl(xl,x2), h2(xi,x2)) where

- a log R(x x )

ax;

13

It can be seen

that this situation exists when the distribution is the product of the

independent exponential marginals .

hi(xi,x2 ) = A+(x.i), ij =1,2; i#j. (2.24)

(2.24) characterizes the Gumbel 's (1960)

bivariate exponential

R(xi,x2) = exp{-21x1- ,2x2-(5xlx2}, (2.25)

A, ,

2 2>0, 0<_8<_212 2.

property of the random vector (X1,X2) by the relations

P(X;>x;+y; IX1> xl, X2>x2) = P(X, > yjl X>>xi), ij =1,2; i$j (2.26)

or

G i(xi+yi,

and

x2) = G1(xl,x2)G1(y1,x2)

(2.27)

yl,y2 >0

(1988b ).

14

by

p(X,>_x,+y,IX,>_ x,, X;=xj) = P(X, >_y,l Xj =x1), ij = 1,2;

(2.28)

for all x,, y, >0 and proved that ( 2.28) holds iff the distribution is the

bivariate exponential by Arnold and Strauss

J(xl,x2) =

0, xl,x2 >0

(2.29)

where 0 =0(8) 8e""8 /-E,(1/8)with E,(u)

W

f e-'"w-ldw

u

property in the bivariate continuous case , a similar approach can be

taken in the discrete case also . The bivariate discrete Lack of Memory

Property in the usual case is

P(Xi>

X12t, X22!t)=

P(Xl>

xl x2 ,t =0,1,2,... .

15

R(x i ,x2) _

P " Pi

i -xz

P xz P22

x1

(2.31)

P(X1 = X1 X2 = x2)

l+p?p1+P2; 0<p<_P;<1, p = pl+p2-c3-1 with c3 =

P(X1 X1 X2 = x2)

P(X,? x,+t IX1>xl, X2>_x2) = P(X1 ?tI Xi>_xi) (2.32)

for all xl, x2, t = 0,1, 2, .,. is characteristic of the bivariate geometric

distribution with survival function given in Nair and Nair ( 1988a).

The conditional LMP in the discrete case is studied by Nair and Nair

(1991). For multivariate results both in the continuous and discrete

vectors we refer Puri and Rubin (1974), Puri (1973 ), Shaked et.al.

(1995) and Zahedi (1975).

definition of failure rate defined by Nair and Hitha (1989), Kotz and

Johnson ( 1991) extended the use of partial sums to the bivariate

discrete case. The possible distributions are geometric , waring and

negative hypergeometric.

lifetime of a device or a component or an organism. If the survival

16

byf(x), then the failure rate h(x) is given by

h(x) = f (x)

R(x)

d log R(x)

(2.33)

dx

The mean residual life

r(x) = E(X-xIX>x)

R(x) f R(t)dt

(2.34)

and r(x) uniquely determines the distribution through the relations

x

(2.35)

and

R(x) = r(O) exp - J 1 dt .

r(x) 0 r(t)

(2.36)

of the failure mechanism into higher dimensions is that there is no

unique way of representation. In the bivariate setup (X1,X2) is a nonnegative random vector admitting the continuous distribution function

F(xi,x2). The survival function is denoted by

17

(2.37)

a2 R

.f(x1,x2) =

If (X1,X2)

is

(2.38)

ax l ax2

as

treated A the lives of the components in a two-

(1971) is

a(xl,x2 ) =

,x z)

R(x1,x2)

f(x1

(2.39)

are independent and exponentially distributed .

the survival function as

a(xl,x2 ) =

aH aH a2H

&1 ax2

(2.40)

&1&2

where

H = - log R(xi,x2).

Johnson and Kotz (1975) gives another approach to define the

bivariate failure rate as a vector (hl(x1 ,x2), h2 (xl,x2 )) where

h;(x1,x2) _

-alogR(x1,x2)

&i

, i=1,2. (2.41)

18

uniquely through the equation

x2

R(xi,x2) = exp

{_

h1(tio)dI1 -

or

x2

R(xi,x2) = exp

xi

0

(2.43)

(c1(x1,x2), c2(x1,x2)) where

c,(x1,x2 ) -aR(xXx2) , i,j=1,2; i#j.

(2.44)

ax3

Buchanan and Singpurwalla (1977 ) define the bivariate mean

residual life ( m.r.l.) function r(x1,x2) by

r(x1,x2 ) = E((X1-x1) (X2-x2) 1 X1>x1, X2> x2) (2.45)

The second definition is provided by Shanbhag and Kotz(1987) and

Arnold and Zahedi

function on R2 '= {(x1 , x2), O<x1 , x2<0o } is defined by (rl(x1 , x2), r2(x1,x2))

r1(x1,x2 ) = E(X,-x; I X1>x1, X2>x2)

I f R ( t ,xJ )dt , i , j =1 , 2 ; i #j

R(xl,x2) x,

(2.46)

( 2 . 47 )

19

terms of ri(xi,x2) and r2(xi,x2) as

R(xi,x2 ) =

r,(0,0)rz(x>,0)

ex p

r1(x1,0)r2(xi , x2)

x, dt

X2

dt

-O r1(t,0) - i1x1,t) j

(2.48)

or

R(xi,x2) =

r,(O,x2 )r2(0,0)

- j^ dt 2 dt

exp J J

r,(x,,x2 )r2(0,x2)

0 ri(t,x2) 0

r2(O,t)

(2.49)

are provided

variable having the survival function R(x)= P(X >-x) and probability

mass function f(x). The failure rate h(x) is defined to be

h(x) = f(x)

(2.50)

r(x) = E(X-x I X>x)

1

R(x + 1) r= R(t) .

(2.51)

(Salvia and Bollinger, 1982)

x-1

Y=O

(2.52)

20

and

R(x) =

[

r(u

(2.53)

r(u)

where f(0) is determined such that Yf(x)=1. Hitha and Nair (1989)

x

, = 0,1, ...

1- h(z+1) _ r(x) - lx

r(x+ l)

Life distributions with virtual hazard rate, mean residual life etc.

were studied in Abouammoh(1990) and Roy and Gupta (1992).

have a single quantity for failure rate as provided by Puri and Rubin

(1974) and Puri (1973 ). They define the multivariate failure rate as

a(xl ,..., x ) =

= xn

(2.55)

P X,>_x1,...,Xn>-xn)

a(zi,x2 )

P(X,=xX2=x2)

Xz z)

> x

= p (Xt > x 1,-

(2.56)

(1990) as a vector failure rate (h1(xl,x2), h2(xl,x2)) with

P(X; = x;,X3 >- xi.)

hi(xi,x2) _

21

and this vector determines the survival function uniquely through the

formula

x'

x2

r=1

(2.58)

r=1

Kotz and Johnson (1991) who view it as the vector (c1(x1,x2), c2(xl,x2))

where

c;(xt,x2 ) =

P(X1

x1,X2

P(X; x;, Xj = xj )

(2.59)

al.(1995).

higher dimentions is due to Nair and Nair ( 1988a) in which they define

the m.r.l. in the support of Iz ={(xl,x2): x1,x2=0,1,... } as (rl(xl,x2),

r2(xi,x2))

where

r;(x1,x2 ) = E(X;-x; I X1? x1, X2>_x2).

(2.60)

It is also proved that r;(xl,x2)= c; for i=1,2 if and only if X;'s are

independent geometric random variables and ri (x1,x2 ) = A;(x1), i, j=1,2;

22

function

R(xl,x2) =Pi' Pie

9x'x2,

0_P1, P2_ 1; 0 <_9<

distribution function F(x) with respect to the Lebesgue measure in the

support of the set of non - negative reals and having a finite mean U.

Associated with X, a new random variable Y is defined, whose p.d.f is

(2.62)

f2

has a special significance in the context of renewal theory.

interest . Let us start the experiment with a single new component at

time zero and replace it upon failure by a new second component and

so on . These failure times X;, i=1,2,.. are independent and let they are

identically distributed with distribution function

S, =X1+X2 +...+X will constitute renewal process .

F(x). Then

Take a sampling

23

before t.

Let

VV be the time measured from t to the next renewal to occur after it.

That is VV is the residual life time of the component in use at time t.

Cox (1962) proved that the limiting distribution of Ut and Vr is

common, called the equilibrium distribution and has a density of the

form specified by (2.62). In this physical situation Y represents the

residual life of the component whose life length is X. For the

applications of equilibrium distribution in reliability studies, we refer

Scheaffer (1972), Rao(1985), Deshpande et al. (1986) and Blumenthal

(1967).

utilised to explain the phenomenon of ageing . Gupta (1984 ) obtained

the equilibrium distribution as a weighted distribution with weight

[h(x)]-' where h (.) is the failure rate . Let G (.) denote the survival

function of Y.

with that of the parent distribution in the context of reliability are

studied by Gupta (1984), Gupta and Kirmani (1990) and Hitha and Nair

(1989). Some of the important identites among them are

24

distribution of a unit which ages more rapidly will come off worse in a

comparison of R(x) and G(x). The wide spread applicability of

weighted distribution in univariate case has prompted many researchers

to extend the concept to the higher dimensions .

However the

out.

0<x1,x2<oo} with an absolutely continuous distribution function

F(xi,x2) or the survival function R(x1,x2). Defining w(x1,x2) be a nonnegative weighted function with E[w(X1,X2)] <oo, Mahfoud and Patil

(1982) defined a bivariate weighted distribution as the distribution of

the vectov(Y1,Y2) with p.d.f

g(x1,x2 ) =

w( x1,x2 )f (x1,x2)

E[w(X1,X2)]

When

w(xl,x2) = 1/h(xl,x2)

(2.63)

(2.64)

G(x1,x2) = f f R(tdt,dt2 (2.65)

xI X2 !"

hy(xl ,x2)

= 1/r(xl,x2 )

(2.66)

where hy(.,.) is the scalar failure rate of (Y1,Y2) and r (xl,x2) is the

scalar m . r.l function of (X1,X2).

introduce the concept of setting the clock back to zero property. As

mentioned earlier, this setting the clock back to zero property can be

defined as an extension of the lack of memory property. Rao and

26

zero (SCBZ) property.

under the following three operations, except for

the value of the parameters , that is

f(x, 0)

-^ f(x, 0

*)

(2.67)

where 0 * E O,

1. Truncating the or`ginal distribution at some point xo_0.

2. Considering the observable distribution for life time X ?xo and

3. Changing the origin by means of the transformation given by

X1=X-xo, so that X1 ?0.

restated as the following

is said to

have the SCBZ property if for each xo >_0 and 0 EO, the survival

function satisfies the equation

R(x+xo, 0) = R(xo, 0) R(x, 0 * )

with 0* = 0 *(xo)EO.

(2.68)

27

0

X is said to have SCBZ property if

P(X>_x+xojX>

_xo) = P(X*>_x),

(2.69)

parameters .

changed. The parameters which does not undergo any change under

the SCBZ transformation has been called normalizing constants.

xo and then setting the origin at xo leaves the form

of the distribution

ensures

organism or a device, given that it has survived xo time units is the

same as the unconditional distribution except for a slight change in

the parameters. That is the residual life distribution (RLD)has the

same form as that of the original distribution except for a change in

the parameters.

28

Rao (1990 ) proved that the life expectancy has a simple form

if the family has SCBZ property and used this property of the time to

tumor distribution with survival function

7 7

, a, 8, y>0

Chiang and Conforti (1989). If r(xo,O)

would survive , on the average

r(xo, 0) = EB(X-xo I X>xo)

f r R(x, 0)

dx

S R(xo,0)

f'R(x+xo,0)dx

O R(xo,0)

That is

(2.70)

0

hazard model , the model for time to tumor given

Gompertz,

linear

by Rao

(1990)

29

from any point on it taken as the origin .

failure rate model. Another application of this SCBZ property is

given in Rao et. al. (1993a ). They have shown that the family

of survival distributions under the proportional hazard model and

accelerated life models have SCBZ property if the baseline survival

distributions have .

tumor free life expectancy are also considered there. Rao and

Damaraju (1992) have shown that the inequalities in the definitions

of the measures New Better than

than Used in Expectation

(NBUE)

bivariate case which he called the extended SCBZ property.

and R2 with hypothetical life times Xi and X2 respectively. The joint

survival function of Xl and X2 is defined by R(xl,x2,9) , 0< xl,x2<00,

where 9 is the parameter or a vector of parameters .

The survival

function of the individual upto age xo can have the idea that the

30

The

induvidual due to risk R1 given that the individual has survived for a

time of xo units is

=

P ( X i ? xl + xo I X 1 _xo,

>

X2 _xo)

>

R(x , + x o, x o, O)

R(xo'x0'O)

(2 . 71)

In a similar way,

=

P ( X22!x2 + xo I X l ? xo, X2 _xo)

>

R(xx2 +xo,O)

R(xo, xa, 0)

(2 . 72)

Using this notations Rao et. al.(1993b) defined SCBZ property in the

bivariate case as follows.

is said to have the SCBZ property if for each 0(=-O and xo_!O, the

survival function satisfies the pair of equations

R(x, + xo , xa , O)

R(xo , xa , O)

R(xl,xo, 0 * )

(2.73)

and

R(x'x2 +x'O) = R(xo,x2, O**)

R(xo, xo,O)

where O*= O*(xo) and O** = O**(xo) Oo where Oo denote the boundary

of O.

31

r2(xo,xo, 6)) has a closed form, since

--xo, X2?xo)

ri(xo ,xo,e) = EB(Xi- xo JXl >

1

R(xo,xo,9) 00

where f(x,,x2,O) is the joint p.d.f of (X1,X2). That is

r 1(xo,xo,9)

1R(x, +xo,x0,6)

dx

o

R(x o , x o, 0)

W

= f R(x,,xo,B*)dx, .

0

Similiraliy

r2(xo,xo,e) = f R(xo,x2,6#)dx2

0

Gumbel ( 1960), Bivariate Gompertz and Bivariate Pareto models.

- Discrete Probability DistributionUploaded bydogy64
- Bayesian Methods for HackersUploaded byohdoubleyou
- STAT 3502 Outline Winter 2015Uploaded byBob
- Business Analytics by James Evans TOCUploaded byArchie
- December 09 10Uploaded byRush Kirubi
- Screen ns Lab 5Uploaded byParsa Far
- October 2010Uploaded byMrunmai Gadbail
- Theory of ProbabilityUploaded byMeet
- MATH1151 Course OutlineUploaded byPatricia
- 00 789814730761.pdfUploaded bywilsonlam2003
- 15MA301-U2Uploaded byvikas sharma
- Queing Theory Model.pptUploaded byHarneet Sachdeva
- ch3Harrell Ghosh Bowden cap2Uploaded byHanswen Julca Anaya
- Poisson ProcessUploaded byHitesh Sen
- Risk Avoidance in InsuranceUploaded bysumitjain25
- FinalUploaded byKulsum Jainoodin
- 13a S1 January 2014Uploaded byxafifax
- ps19ansUploaded byantonyluk
- CHAPTER 5 Reviewer (1) (1) (1)Uploaded byLica Falculan
- l3 probability methodsUploaded byapi-287224366
- Section 05 01 Ess Stats2eUploaded bykaled1971
- Lecture 2 - Statistics.pptxUploaded byMohanad Suliman
- Chapter 5_ Discrete Random VariablesUploaded byDelfin Valdez
- The Analysis of Performance Measures of Generalized Trapezoidal Fuzzy Queuing Model with an Unreliable ServerUploaded byAnonymous 7VPPkWS8O
- chap2_9_7Uploaded byAdry Genialdi
- 4703-07-Notes-ARMUploaded byPhuong Hoang
- SimulationUploaded byCamille Santos
- Wk 1 Appendix ReviewUploaded byMahesh Jung Khadka
- 01 Random VariableUploaded byharshasud
- Syllabus Gb 210 Spring 2010Uploaded bygshah9749

- Social PerceptionUploaded byArshiya Jabeen
- Student Retail 8e - Chapter 10.pptUploaded byArshiya Jabeen
- MM SyllabusUploaded byArshiya Jabeen
- Nature, Scope and Challenges of Industrial MarketingUploaded byJohn See
- Basic Concepts of Marketing 120129183532 Phpapp01Uploaded byArshiya Jabeen
- Marketing Information System (1)Uploaded byArshiya Jabeen
- QuizUploaded byArshiya Jabeen
- Retail CommunicationsUploaded byArshiya Jabeen
- BS QP modelUploaded byArshiya Jabeen
- Project TitlesUploaded byArshiya Jabeen
- Sudhir ProjectUploaded byArshiya Jabeen
- T&D in IndiaUploaded byArshiya Jabeen
- T&D in IndiaUploaded byArshiya Jabeen

- Stat 2008Uploaded bykobe
- Edu 2013 Fall Mlc SyllabusUploaded byMalik Amin
- centralLimitTheroem.pdfUploaded byARIMANDA RAVI TEJA REDDY
- ch4Uploaded bySpencerHarrison
- The Box PlotUploaded byIwan
- Parzen 1960Uploaded byVictor Vargas
- edu-2009-spring-p-quesUploaded bycacarroto
- Condorcet Meets BenthamUploaded byNam
- Multivariate Random VariablesUploaded byEdmond Z
- Probability DistributionUploaded byHayati Aini Ahmad
- Fizionometria umanaUploaded bySescioreanu Mihai
- Syllabus 380 134Uploaded bychou_timothy
- 6SEM SYLABUSUploaded by9chand3
- (Courant Lecture Notes in Mathematics 7) S. R. S. Varadhan-Probability Theory-Courant Institute of Mathematical Sciences_ American Mathematical Society (2001)Uploaded byRodrigo
- Chapter 8 Binomial DistributionUploaded byBid Hassan
- DIS_ch_7.pptxUploaded byVictoria Liendo
- LxMLS2015Uploaded byyaqian
- part_iiUploaded byJacob Harris
- Chapter 02 answers.pdfUploaded byadow
- Sample MomentsUploaded byAshish Sachan
- mss11 statistics syllabus - course outlineUploaded byapi-264533666
- Discrete Distribution 1Uploaded byNeelam Yadav
- Cape Integrated Mathematics - SyllabusUploaded byRobert Edwards
- STAT3004 Course NotesUploaded byShivneet Kumar
- Sample Questions for Midterm IIUploaded byDuy Do
- Remind Aleatory Variability and Epistemic UncertaintyUploaded byXianju Yiang Li
- 04 Point EstimatesUploaded byRangzaib Munir
- b Da Lecture NotesUploaded byxolrax
- 5Uploaded byKowsik_JS
- Statistics and Probability - Midterm Examination (TOS)Uploaded byRaymond Richard Magan