TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.6, Jun
e
201
4, pp. 4617 ~ 4
6
2
3
DOI: 10.115
9
1
/telkomni
ka.
v
12i6.543
8
4617
Re
cei
v
ed
De
cem
ber 2
9
, 2013; Re
vi
sed
F
ebruary 28,
2014; Accept
ed March 1
5
, 2014
A New Statistical Mod
e
l to Estimate Information System
Contingency Budget
J
u
n
w
ei Z
e
ng*
1
, Fajie Wei
1
, Haitao Xiong
2
, An
y
i
ng Liu
1
1
School of Eco
nomics a
nd Ma
nag
ement, Bei
han
g Univ
ersit
y
,
Beiji
ng 1
0
0
191
, China
2
School of Co
mputer an
d Informatio
n
Engi
n
eeri
ng,
Beij
in
g T
e
chnolog
y
an
d Busin
e
ss Uni
v
ersit
y
,
Beiji
ng 1
0
0
048
, China
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: zengj
un
w
e
i@
ye
ah.n
e
t
A
b
st
r
a
ct
Development
of an
infor
m
ation system
is a c
o
mple
x pr
oces
s, which ex
pos
e to
a
great
num
b
er
of
risks.Henc
e, the high failure r
a
tes
long ass
o
ciated with
inform
ation system
projects, despite
adv
ances
in
techni
qu
es for infor
m
atio
n techn
o
lo
gy dev
elo
p
m
ent, sug
gest that org
a
n
i
z
a
t
i
ons n
eed
to impr
ove th
e
i
r
abil
i
ty to identif
y and to mana
ge assoc
i
ate
d
risks. To
impro
v
e the risk ma
nag
e
m
ent i
n
in
formati
on syst
e
m
deve
l
op
ment p
r
ojects, a pra
g
m
atic
proce
dur
e is sug
geste
d to deter
min
e
th
e si
z
e
of a pr
oj
ect'
s conting
e
n
c
y
pla
n
b
u
d
get
at any
spec
ifie
d
level
of c
e
rtai
n
t
y. Consi
deri
n
g
the
intera
ction am
ong r
i
sk factors, a
m
e
thod
base
d
on c
o
mmo
n
risk factor
s and co
pu
la f
unctio
n
s is
us
ed to
mo
del
a
nd q
uantify p
o
s
itive de
pe
nd
e
n
ce
betw
een risks.
Ke
y
w
ords
:
infor
m
ati
on ex
traction, copu
l
a
, informatio
n
system deve
l
op
me
nt projec
t, contingency
pla
n
bud
get, risk de
pen
de
ncy
Copy
right
©
201
4 In
stitu
t
e
o
f
Ad
van
ced
En
g
i
n
eerin
g an
d
Scien
ce. All righ
ts reser
ved
.
1. Introduc
tion
Even before
the wide
sp
re
ad use
of “the mythical m
an-m
onth”
[1]
,
information
system
(IS) p
r
oje
c
t failure
was
de
scribe
d a
s
a
co
mm
on
p
h
e
nomen
on an
d
many devel
opment proj
e
c
ts
did n
o
t a
c
hie
v
e previously
co
st,
schedu
le, and
pe
rformance
goal
s. It ha
s b
een
continuin
g
to t
h
e
pre
s
ent. Ove
r
last three
decade
s, the
r
e ha
s b
een
con
s
ide
r
a
b
l
e
intere
st in
explorin
g a
n
d
explaining th
e rea
s
o
n
s fo
r the abno
rma
l
high failure rates. Ma
cF
a
r
lan p
o
ints o
u
t that failure
to
asse
ss
indivi
dual proj
ect ri
sk
is
a majo
r sou
r
ce
of
the
IS develop
m
ent p
r
oble
m
[
2
]. Then, m
a
n
y
resea
r
chers,
therefo
r
e, try
to identify th
e vario
u
s ri
sks a
s
so
ciated
with the IS
d
e
velopme
n
t [3].
Jian
g et al. argue that
the high failure rates
asso
ciated
with IS projects
sugg
est
tha
t
orga
nization
s need
to im
prove
not o
n
ly their
a
b
il
ity to identify, but also
to mana
ge t
h
e
asso
ciated ri
sks [4]. Also, Project ma
nagem
ent re
sea
r
che
r
s
state that risks in informat
ion
system
devel
opment
proje
c
ts are
key f
a
ctors affecti
ng p
r
oj
ect
su
ccess [5].
Fo
r exam
ple,
Di
llon
et al. sho
w
the impo
rtan
ce of ri
sk m
anag
em
ent
and
conting
e
n
cy re
se
rve
s
for su
cce
s
sfu
l
developm
ent
of p
r
oje
c
ts in
com
p
lex
and
un
ce
rt
ain e
n
viron
m
ent [6]. Tha
t
re
sea
r
ch a
l
so
recogni
ze
s th
e ch
alleng
es
of managi
ng
these
re
se
rve
s
.
More
and more re
sea
r
ches reveal
th
at,
to mana
ge t
he in
herent risks
asso
ciat
ed
with
dyna
mic e
n
viron
m
ents
and
u
n
ce
rtaintie
s
(e.g.
requi
rem
ent chang
es in IS proje
c
ts), eno
ugh re
so
urce
s sh
ould b
e
a
ssi
gne
d to project
s
[7-9].
A co
ntingen
cy plan i
s
de
si
gned
for the i
dentif
ied
un
certainty, which me
ans that
peo
ple
can have a relatively controllabl
e environmental
se
tting. This feat
ure is very m
eanin
g
ful for the
IS developme
n
t proje
c
ts
be
cau
s
e m
o
st o
f
t
he cha
nge
s can
be p
r
edi
ctable
and
project ma
nag
e
r
s
have a
lots
of alternative tech
nolo
g
ical tool
s to
redu
ce
or co
nstrai
nt the
i
m
pact
of tho
s
e
predi
ctabl
e chang
es. Som
e
proj
ect ma
nage
rs t
r
y
to use a fixed
determi
nistic percenta
ge
of
proje
c
t bud
g
e
t to capture
and quantif
y the degree
of confiden
ce that the continge
ncy pl
an
sho
u
ld cover.
Toura
n
critici
z
e
s
the use o
f
this appro
a
ch [10].
Budgeting fo
r proj
ect
conti
ngen
cy pla
n
, an efficie
n
t way to re
serve
re
sou
r
ces, h
a
s b
e
e
n
studie
d
by scholars [10
-
12
]. Rece
ntly, Khamoo
shi
an
d Cioffi pro
p
o
s
e a p
r
ag
mat
i
c procedu
re
to
determi
ne the size of co
n
t
ingen
cy plan
budget fo
r a
proje
c
t, a progra
m
s, or a
system at any
spe
c
ified
lev
e
l of
ce
rtaint
y confid
en
ce
[8]. T
hat m
e
thod, li
ke
many oth
e
r
risk ma
nage
ment
approa
che
s
, con
s
id
ers ri
sk
s as in
dep
en
dent events.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4617 – 4
623
4618
Ho
wever, K
w
an an
d Le
un
g argue th
at hypothe
sis i
s
cou
n
terint
uitive as it i
s
mo
re li
kely
that one
ri
sk
woul
d imp
a
ct
anthe
r
risks
[13]. This
is reasona
ble, e
s
pe
cia
lly i
n
ri
sk ma
nage
m
ent
of IS develo
p
ment
proj
ects. So, the
indep
ende
nce
assu
mption
betwe
en
th
e
s
e ri
sks ca
n
be
spe
c
io
us [1
4
]. The long
-standi
ng i
s
sue of
de
pen
den
ce b
e
twe
en ri
sks
ha
s re
cently be
en
discu
s
sed
i
n
proje
c
t risk
a
nalysi
s
[15-1
7
]. These re
searche
s
un
a
n
imou
sly model depe
nde
nce
usin
g the co
p
u
la app
roa
c
h
[18, 19].
In this re
se
arch, the
comb
ination
of ri
sk
tole
ra
nce
a
nd statistical depe
nden
ce
will
b
e
analyzed a
n
d
then b
e
u
s
e
d
to allo
cate t
he conting
e
n
c
y plan
bud
g
e
t in IS devel
opment
proj
e
c
ts.
A applied
proce
dure is propo
sed to
m
anag
e the
co
ntingen
cy pla
n
bud
get of I
S
developm
e
n
t
proje
c
ts, co
nsid
erin
g
de
pend
en
cies among ri
sk events in
IS projects. The pro
c
e
dure
demon
strates ho
w to fo
rm
ulate a
proje
c
t contin
g
e
n
cy plan a
nd t
o
allo
cate
co
ntingen
cy pla
n
budg
et for risks d
e
fined ov
er the duratio
n of an
IS de
velopment project. Taki
ng
the cou
p
ling
of
blocks in
a i
n
formatio
n
system into
a
c
count, th
e
copula
is intro
duced to
mo
del relation
ship
betwe
en d
e
p
ende
nt risks t
o
ma
ke the
e
s
timation m
o
re practi
cal. T
he Mo
nte Ca
rlo meth
od
s
are
also u
s
e
d
in the cal
c
ul
ation
of t
he joint distributio
n of ri
sk eve
n
ts.
In the re
st of this pap
er, a statistical
depen
den
ce
model for I
S
proje
c
ts i
s
firstly
descri
bed. A
nd then, a framewor
k that
enable p
r
oje
c
t mana
gers
to work bette
r with inevita
b
le
risks in p
r
oj
ects is presente
d
. T
he last se
ction is the
co
nclu
sio
n
.
2. A Statis
tic
a
l Depen
d
en
ce Model for
Information
Sy
stem
2.1. Method
s
Curren
t
ly
U
sed
Cioffi et al. [2
0] present a tool that helps
manage
rs to deal with “ta
c
tical ri
sks” which a
r
e
defined i
n
[21
]. Given a sp
ecified l
e
vel o
f
certai
nty, e.g. 99\%, their pra
g
matic proce
dure
can
be
use
d
to
determine the
si
ze
of a
proj
ect'
s risk
co
ntinge
ncy b
udg
et. Acco
rdi
ng to
their d
e
scripti
on,
the key ideal
of the pro
c
ed
ure
can b
e
summari
ze
d a
s
: use
binomi
a
l distri
bution
to estimate the
prob
ability of occurre
n
ce of any spe
c
ific
numbe
r of ri
sk events h
a
p
penin
g
and th
en cal
c
ul
ate the
size of the potential dam
age co
rrespo
nding with a
given confid
ence level. More
spe
c
ifically,
con
s
id
erin
g e
a
ch i
ndividu
al risk eve
n
t m
a
y or may
no
t occur, p
eopl
e ca
n represent all po
ssib
le
scena
rio
s
by
combi
nation
s
of these in
dividual ri
sk events. Finall
y
, they sugg
est two diffe
rent
ran
k
ing
sche
mes, sortin
g
risk event
s
by their
imp
a
cts
or thei
r expecte
d value
s
, to set
the
contin
gen
cy plan bu
dget.
They a
s
sume
that e
a
ch ri
sk eve
n
t i
s
tot
a
lly
inde
pen
d
ent of th
e oth
e
r
risk event
s and
the
increa
se i
n
i
m
pact
be
cau
s
e
of po
ssibl
e
compo
undi
ng effe
cts
co
uld b
e
in
clu
d
ed by
agg
reg
a
ting
con
n
e
c
ted, subsequ
ent i
m
pact
s
into f
e
we
r, la
rge i
m
pact
s
. So,
the variabl
es that sho
u
ld
be
con
s
id
ere
d
are the total number of po
ssi
b
le risk
event
s, the averag
e risk
proba
bi
lity,
the numb
e
r
of risk eve
n
ts that will o
ccur, the give
n
conf
id
en
ce le
vel of any sp
ecific
ri
sk
out
come,
and th
e
impact of the
risk events.
Although extremely novel proje
c
ts fa
ce
risk ev
e
n
ts f
a
r out in the tails of a pro
bability
distrib
u
tion, they point o
u
t
t
hat most
orga
nization
s do not fa
ce
these
risk
events o
n
m
o
st
proje
c
ts
and
work o
n
proje
c
ts o
n
ly slight
ly differ
ent fro
m
the one
s th
ey worke
d
on
before. Thi
s
i
s
esp
e
ci
ally true for IS development proj
ects. Fo
r
mo
st of the IT compani
es, co
de reu
s
e i
s
ver
y
popul
ar. In some large co
mpanie
s
, alm
o
st all ne
w in
formation
systems are dev
elope
d by usi
n
g
the existing software a
r
chit
ects a
nd/or
software d
e
vel
opment platfo
rms.
The probabilit
y of occurrence
of
each
risk event as well as
its
im
pact can be estimated
by many ri
sk
manag
eme
n
t pro
c
ed
ures.
These e
s
tima
tion pro
c
e
d
u
r
es a
r
e
com
m
on practi
ce a
nd
the results are often given as point esti
mates.
Peopl
e also can ca
lculate the ex
pecte
d value of
the loss by m
u
ltiply the pro
bability and t
he lo
ss
(u
s
u
a
lly written a
s
,
[22]). Both “P
” an
d “I”
are
the
basi
c
inp
u
ts
of the pro
c
ed
ure an
d the a
c
cura
cy is lo
w be
cau
s
e g
ood ri
sk mod
e
ls an
d ha
rd
data
are rarely ava
ilable.
Then, [20] try
to use
statist
i
c metho
d
s to
make
app
rox
i
mation of an
averag
e o
c
cu
rre
nce
prob
ability of risks after th
ey sho
w
ing t
he pro
bab
ility
of occu
rren
ce of m
any risk events at t
h
e
same
time i
s
low. O
n
ly after g
e
tting tha
t, can the
bin
o
mial di
strib
u
t
ion be
used
to estimate
the
numbe
r of risks that can o
c
cur at
any gi
ven confid
en
ce level.
There are two main drawbacks to get the
accuracy
and relia
ble
result
s by u
s
ing the
above proced
ure in the IS developm
ent proje
c
ts:
(1) A critical
assumption
in [20], the i
ndep
ende
nce
betwee
n
ri
sk events
whi
c
h is a
comm
on p
r
a
c
tice
of current ri
sk
man
ageme
n
t app
roa
c
he
s, i
s
d
i
scusse
d by
more
and
m
o
re
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A New Statist
i
cal Mod
e
l to Estim
a
te Info
rm
ation Syste
m
Contingen
cy Bud
get (Ju
n
wei Ze
ng
)
4619
resea
r
che
s
. Among the
s
e re
sea
r
che
r
s, [17] state
s
that it is intuitively obvious th
at the
assumptio
n
is highly suspect for m
a
ny large
p
r
o
j
ects
whi
c
h
may contai
n
multiple si
milar
activities o
r
several diffe
re
nt kind
s of
a
c
tiviti
es that
can
be influ
e
n
ce
d by com
m
on ri
sk fact
ors.
[13] arg
u
e
s
th
at the assu
m
p
tion is
co
unt
erintuitive be
cau
s
e it i
s
mo
re li
kely that risks in
one
area
woul
d impa
ct risks in a
noth
e
r area.
(2) Althoug
h
it is m
u
ch e
a
s
ier to
estim
a
te the
prob
ability of o
c
curren
ce
and
damag
e
sep
a
rately, th
e limitation of
the stati
c
e
s
timati
on (m
ost of the, point
estimate
s) o
f
damag
e/loss
can n
o
t be ignore
d
be
cau
s
e the cal
c
ulat
ion of t
he potential dama
g
e
is ba
sed on
the value from
risk id
entifica
t
ion an
d q
u
a
n
tification. In
pra
c
tice, th
e dam
age
caused
by ri
sk eve
n
ts
ca
n
be
several different levels of severity
with different occu
rrence pro
babil
i
ty.
In the
belo
w
sub
s
e
c
tion,
a
statisti
cal
de
pend
en
ce m
odel i
s
propo
sed
to
deal
with the
first d
r
a
w
ba
ck a
nd a
ne
w
pro
c
ed
ure wil
l
be d
e
scribe
d
in n
e
xt se
ction to ove
r
come the
second
dra
w
ba
ck.
2.2. A Copul
a-ba
sed Sta
t
istical Dep
e
ndenc
e Mod
e
l
The imp
o
rta
n
t
of relaxing
the ind
epe
nd
ent assu
mpti
on ha
s
bee
n
clea
rly re
co
g
n
ize
d
by
s
c
h
o
l
ar
s
.
Th
e
c
o
pu
la
ap
p
r
oa
c
h
is
us
ed
to
s
o
lve
thi
s
i
s
sue by m
a
n
y
research
ers [15-1
7
], [23]. A
cop
u
la i
s
a f
unctio
n
that l
i
nks the ma
rginal di
strib
u
tions to th
e j
o
int distri
buti
on, whi
c
h i
s
a
statistical con
c
ept that rela
tes ra
ndo
m variabl
es
.
Ref
e
ren
c
e [1
7] d
i
scusse
s a
si
mulation
-ba
s
ed
model to qua
ntify positive
depe
nden
ce
betwe
en un
certai
nty distri
bution
s
of activities in a project
netwo
rk.T
heir model can p
r
ovide a le
ss cumb
ersom
e
method to
elicit dep
end
ency info
rmat
ion
from expe
rts.Their
re
sea
r
ch i
s
u
s
eful
in
un
ce
rtain
t
y analysis
whe
r
e d
epe
n
den
ce b
e
twe
en
rand
om va
ria
b
les
with
a b
ound
ed
sup
p
o
rt is pr
esent
due to
co
m
m
on fa
ctors.
Later, [16]
p
u
ts
forwa
r
d
a mo
del for b
u
ildin
g statisti
cal d
epen
den
ce
b
e
twee
n ma
rgi
nal di
stributio
n.Wu et al. [2
3]
employ multivariate copul
a to model
the depe
nden
ce
among
risk fa
ctors.
The
co
re i
d
e
a
of u
s
in
g
copula
metho
d
to
cop
e
wit
h
de
pen
den
cy is th
e p
r
o
c
edure to
cal
c
ulate the
multivariate
distributio
ns betwe
e
n
subsets of u
n
ce
rtainty distributio
ns.After
analyzi
ng the
methods to d
eal with two e
x
tremes
cau
s
ed by assumi
ng the margi
n
al distrib
u
tions
to be
sp
ecifie
d sepa
rately,
the multivari
a
te ca
n b
e
co
nstru
c
ted
by
an inte
rme
d
ia
te metho
d
. T
he
method a
s
sumes in
dep
ende
nce betwee
n
margi
nal distri
buti
ons a
nd all
o
ws to use
joint
distrib
u
tion
s f
o
r
sub
s
et
s of
un
certai
nty distrib
u
t
i
on
s whi
c
h sha
r
e
comm
on ris
k
f
a
ct
or
s.
A
c
t
u
ally
,
the ideal of latent variabl
e model
s ha
s bee
n
used
when
spe
c
if
ying the inde
pend
en
ce ([2
4
,
25]).Peopl
e can use many skill
s,
e.g. bra
i
nstorming
se
ssi
on
s, to
identify the common ri
sk fa
ctors
in a spe
c
ific proje
c
t. The depen
den
ce diagrams
can al
so be
introdu
ced i
n
to proje
c
t risk
analysi
s
.Con
side
ring the
constructio
n
of a rank
co
rrel
a
tion by proje
c
t expert
s
is i
m
pra
c
tical.
The
Diag
onal
Band
dist
rib
u
tion introdu
ced by
[2
6] is
sug
g
e
s
ted to
be
used u
n
d
e
r thi
s
environ
ment
[17]. A bivariate Diag
ona
l Band dist
ri
bution
(,
)
D
UV
of two unifo
rm
on [0,1]
distrib
u
ted ra
ndom
va
ria
b
l
e
s
U
and
V
is
sh
o
w
n i
n
Fig
u
re
1. To m
odel
the
statistical
depe
nden
ce
betwe
en
risk events i
n
a
spe
c
ific
proje
c
t, a multiv
ariate dist
ributi
on of
risk ev
ents n
eed
to
be
modele
d
. A copula
-
ba
se
d statistical dep
ende
nce mod
e
l is pro
p
o
s
e
d
in this pap
e
r
.
Figure 1. A Model for Statistical De
pen
de
nce
of Risks d
ue to Comm
on Ri
sk F
a
cto
r
s
Figure 2. A Diagon
al Band
Distri
bution S
een
from Above [17]
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4617 – 4
623
4620
In Figure1,
Ri
sk
i
is u
s
ed
to de
note a
spe
c
if
ic ri
sk event
who
s
e
potent
ial loss follo
ws
some
kind
s o
f
distribution
and
j
F
actor
repre
s
e
n
ts a com
m
o
n
factor that can imp
a
ct o
n
several
risk events in
a IS development proj
ect.
To e
s
timate
the p
o
tential l
o
ss of
a IS
project, a
multi
v
ariate j
o
int d
i
stributio
n,
0
()
F
, can
be sp
ecifie
d betwe
en
12
3
4
,,
,
R
isk
R
i
s
k
R
isk
R
isk
and
5
R
isk
.In most ca
se, the margi
nal distri
butio
ns,
denote
d
by
()
i
ri
FR
i
s
k
,of
12
3
4
,,
,
R
isk
R
i
s
k
R
isk
R
isk
and
5
R
isk
a
r
e
available
(from
estimation
of
project
engin
eers
an
d expe
rts). B
y
introdu
cin
g
new comm
o
n
influential
fa
ctors
(co
mmo
n ri
sk fact
ors
or
so-call
ed lat
ent variabl
es), the inde
p
ende
nt com
m
on ri
sk factor
1
F
actor
and
2
F
actor
(whose
margi
nal di
stribution can b
e
denote
d
by
()
j
f
j
FF
a
c
t
o
r
), joint dis
t
ribution of five ris
k
events
,
0
()
F
,
can b
e
simplif
ied.
It can be
rep
r
ese
n
ted by th
e pro
d
u
c
t of two ind
epe
nd
ent distri
butio
n
12
3
1
2
3
(,
,
)
F
Ri
sk
Ri
sk
Ri
sk
whi
c
h me
an
s the joint di
stribution fo
r
12
3
,,
R
isk
R
isk
R
i
s
k
and
4
545
(,
)
F
Risk
Risk
whi
c
h me
ans th
e joint
distrib
u
tion f
o
r
4
R
isk
and
5
R
isk
.So, to get the t
w
o joi
n
t di
stribution
s
,
12
3
1
2
3
(,
,
)
F
Ri
sk
Ri
sk
Ri
sk
and
4
545
(,
)
F
Risk
Risk
, the joint distributio
n bet
wee
n
j
F
actor
and
i
R
isk
(j=1, i
=
1, 2, 3 or j=2, i=4
,
5)
sho
u
ld be sp
ecified. The
condition
al
in
d
epen
den
ce
can b
e
u
s
e
d
t
o
si
mplify the
joint
distri
but
ion
to a
com
b
inat
ion of
seve
ral
bivariate
di
st
ribution
s
[1
6]. According
to
the d
e
finition
of
copul
a, th
e
joint distrib
u
tion betwe
en
j
F
ac
tor
and
i
R
isk
su
ch as
1
F
actor
and
1
R
isk
ca
n be u
n
iquely dete
r
mined
by its associ
a
t
ed cop
u
la.
In Figu
re
2,
the DB
copul
a,
(,
)
D
UV
, is a
biva
riate Di
ago
nal
Band
di
strib
u
tion of t
w
o
uniform
on
[0, 1] di
strib
u
t
ed rando
m
variable
s
(
U
an
d
V
) with one
para
m
eter
(
).
The
DB
cop
u
la
can
be u
s
ed to
descri
be the
relation
shi
p
s of ri
sk ev
ents in
Figu
re 2. Un
ce
rta
i
nty
informatio
n
can
be
elici
t
ed and
the
margi
nal
cumulative di
stributio
n fu
nction
of
j
F
ac
tor
,
()
j
f
j
F
Fact
or
and the
ma
rginal di
strib
u
tion fun
c
tion
of
i
Risk
,
()
i
ri
F
Ri
sk
also can
be
figu
red o
u
t.
()
j
f
j
F
Fact
or
and
()
i
ri
F
Ri
sk
contain
all informatio
n on the ma
rg
inal distri
b
u
tion.
From stand
ard
distrib
u
tion th
eory, all marginal dist
ribut
ion wh
i
c
h is
absolute co
ntinuou
s may be derive
d
from
the unifo
rm
margi
nal by
an ap
pr
opri
a
te tran
sformat
i
on.So, both
()
j
f
j
F
Fact
or
and
()
i
ri
F
Ri
sk
are
r
e
la
te
d to
un
ifo
r
m
r
a
nd
om va
r
i
ab
le
s
o
n
[0
, 1
].Hen
c
e
, in
F
i
g
.
2,
U
can
be
asso
ciated
with
()
j
f
j
F
Fact
or
and
V
can re
prese
n
t
()
i
ri
F
Ri
sk
.
Gene
rally
sp
eaki
ng, if con
s
ide
r
ing
the
depe
nden
ce
of risks, the j
o
int dist
ributi
on of ri
sk
events,
i
Risk
, sh
ould
be
obt
ained.Th
e
problem
can
be
simplified
by introdu
cing
som
e
indep
ende
nt comm
on ri
sk
factors,
j
F
ac
tor
, and
usin
g a theo
ry calle
d con
d
itional ind
e
p
ende
nce
[27]. So, only
the bivariate
distrib
u
tion
s of
j
F
ac
tor
and
i
Risk
need to be co
nsi
dered.
The Monte
Carlo meth
od for asse
ssing
variability and
unce
r
tainty in proje
c
t ri
sk
analysi
s
has b
e
come
more
comm
o
n
[28]. The reverse of proce
dure de
ri
ving cop
u
la
s can b
e
u
s
e
d
to
gene
rate pseudo
-rando
m
sam
p
le
s from
g
ene
ra
l cl
asse
s of multivariate
prob
ability
distrib
u
tion
s.That is, give
n a p
r
o
c
ed
ure to ge
ner
ate
a sample
fro
m
the copul
a
distri
bution,
the
requi
re
d sam
p
le ca
n be co
nstru
c
ted.M
o
re details
can
be found in [2
9].
3.
Ne
w
Fram
e
w
o
r
k for
Bu
dgeting
The
pro
c
e
dure p
r
op
ose
d
i
n
this pap
er in
herit
s the
the
o
retically
sou
nd fou
ndatio
n
s
of
the
ran
k
co
rrelati
on method a
llowing the
margi
nal di
st
ribution
s
to b
e
spe
c
ified separately an
d is
practical enough to be used by IS
project analysts.It contains five
steps, as shown in Figure 3.
Risk Id
entifi
cation
: It h
a
s
be
come
the con
s
en
sus th
at the
req
u
irement
of ri
sk
identificatio
n
is to dete
r
mi
ne whic
h ri
sk events a
r
e li
kely to affect
the proj
ect a
nd to do
cum
ent
the cha
r
a
c
teristics of ea
ch
potential ri
sk ev
ent. Besi
de the no
rm
al req
u
irem
e
n
ts, the com
m
on
risk fa
ctors,
e.g., comm
o
n
code
blo
cks, devel
opm
e
n
t frame
w
o
r
k et al., a
r
e
al
so
nee
ded
to
be
identified by experts
an
d
t
hen risk
even
ts a
r
e g
r
o
upe
d a
c
cordi
ng t
o
the
com
m
o
n
ri
sk facto
r
s.The
final result of this step
can
be illustra
ted
by figures
si
milar to Figure 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A New Statist
i
cal Mod
e
l to Estim
a
te Info
rm
ation Syste
m
Contingen
cy Bud
get (Ju
n
wei Ze
ng
)
4621
Risk Q
u
an
tification
: In t
h
is
stage, ri
sks
and i
n
teraction
s
am
o
ng them n
e
e
d
to be
evaluated to
asse
ss the rang
e of possible p
r
oje
c
t outcom
e
s. T
he potential i
m
pact an
d the
probability of occurrence
of each
risk
should
be
evaluated. Usually,
a risk
events can cause
damage at several different levels with corres
pondi
ng probabilities. The
m
o
re scenarios
with
probabilities can be listed in th
is step, the more accurate the final result can be.
Risk Fi
tting
:
It is noticeabl
e that, in thi
s
new
procedu
re, peopl
e
can
take
all un
ce
rtainty
factors into
con
s
id
eratio
n
instea
d of
dealin
g with
impact
and
occu
rre
nce
sep
a
rately
and
determi
nisti
c
ally. In other words, this
pro
c
ed
ure ca
n eliminate t
he se
co
nd d
r
awba
ck of t
h
e
curre
n
tly use
d
method
me
ntioned i
n
se
ction 2.1.
Ba
sed on th
e sce
nario
s from p
r
eviou
s
ste
p
, the
distrib
u
tion
s
of the ri
sk
events
can be simulate
d.
Th
ere are a
lot
s
of
tool
s ca
n
help peo
pl
e
to
transl
a
te the
scena
rio
s
o
f
a risk
eve
n
t into a
sui
t
able di
strib
u
t
ion.The u
n
iform, tria
ngul
ar,
binomial, P
o
i
s
son, exp
one
ntial, Student'
s
t a
nd
nor
m
a
l are ve
ry p
opula
r
di
stri
b
u
tions in
proj
ect
manag
eme
n
t and
mo
st of
proj
ect
man
agers
are fa
miliar
with th
ese
di
stributi
ons. T
he
qu
ality
assuran
c
e (QA) team
s in most of IT compa
n
ie
s are u
s
ing
so
me efficient
bug man
age
ment
softwa
r
e which can h
e
lp to gene
rate the
reli
abl
e asse
ssment of the
distrib
u
tion
s.
Risk Upda
ting
: This
step
is used to u
pdate the di
st
ri
bution
of ea
ch ri
sk in a IS proje
c
t.
The info
rmat
ion extra
c
tio
n
and th
e fitting of
risk events d
o
not co
ntain
the qua
ntifiable
depe
nden
cy i
n
formatio
n. T
h
is
de
sign
is
rea
s
on
able
b
e
ca
use the
i
m
pact
of
com
m
on
risk fa
ct
ors
are n
o
t easy to be accu
rate
ly valued and
the best prac
tice of this wo
rk i
s
to build
a spe
c
ial tea
m
with exp
e
rts i
n
many fiel
ds.With the m
o
del in
se
ct
ion
2.2, the
de
scriptio
n of
distribution
s
of t
he
comm
on ri
sk
factors
can b
e
avoide
d. So, the join
t di
stributio
n of e
a
ch
risk eve
n
t
s gro
up
can
be
cal
c
ulate
d
by con
s
id
erin
g
the dep
end
en
cy stru
cture.
Risk Cumulation
: The a
m
ount of da
mage that th
e IS proje
c
t will experi
e
n
c
e depe
nd
s
on ri
sk
even
ts that will o
c
cur, which
peopl
e ca
nn
ot kno
w
spe
c
ifically in a
d
vance. But, the
expectatio
n
or a
spe
c
ific value given
any leve
l of
certai
nty, e.g., 99\% con
f
idence, ca
n
be
cal
c
ulate
d
after analy
z
ing t
he joint distri
bution of all ri
sk eve
n
ts.
T
h
is
p
r
oc
ed
ur
e
is
ve
r
y
p
r
ag
ma
tic
.
It s
i
mp
lif
ies the
extractio
n
p
r
o
c
e
s
s of expe
rt
opinio
n
without lo
ss o
f
accu
ra
cy.
Figure 3.
Procedur
e for Getting Co
ntin
genc
y Pla
n
Bud
get
4. Conclusio
n
Failures
of IS developm
ent
proj
ect
s
(co
s
t overru
ns,
sche
dule
dela
y
s, poo
r qu
ali
t
y) can
cause enorm
ous losses. Unfortunat
ely
,
the rate of occurrence
of failures i
s
still high. Many
schola
r
s t
r
y to solve thi
s
p
r
oble
m
by software
engi
ne
ering
and
pro
j
ect ri
sk man
ageme
n
t. At the
same tim
e
, continge
ncy pl
an bud
get for IS developm
ent proj
ect, which i
s
a very
import tool f
o
r
proje
c
t man
a
gers to redu
ce the risk exposure, a
ttra
c
ts mo
re an
d
more re
se
ar
che
r
s. H
o
w
e
v
e
r,
there a
r
e two
wea
k
point
s in existing m
e
thod
s:
(1) T
he assu
mptio
n
about ind
e
pend
en
ce of risk
events,
whi
c
h is o
b
viousl
y
cou
n
teri
ntu
i
tive; (2)
The
dete
r
mini
stic d
e
scription
of ri
sk
eve
n
ts,
whi
c
h may cause data lo
ss. The
statistical mo
del and procedu
re prop
osed in this pape
r can
improve
the
accuracy
of t
he e
s
timation
of
cont
ing
e
n
c
y pla
n
b
udg
et for IS
deve
l
opment
proje
c
ts
by overcomin
g
the above t
w
o di
sadva
n
tage
s. T
he five step
s can
help sta
k
e
hol
der to g
r
ab a
nd
extract inform
ation abo
ut ri
sk eve
n
ts in t
he IS pr
oje
c
ts and tran
slate
them into a specifi
c
num
b
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4617 – 4
623
4622
whe
n
give
n a
level of
cert
ainty. Diag
on
al Ban
d
di
stri
bution
and
M
onte
Ca
rlo m
e
thod
s a
r
e
u
s
ed
to quantify the depe
nde
ncy among ri
sks.
Ackn
o
w
l
e
dg
ements
This re
se
arch
was pa
rtiall
y
sup
porte
d by
the Nation
al Nat
u
ral
Science Fo
und
ation of
Chin
a
(No.
7
1201
004,
71
1011
53) an
d
the S
c
ientifi
c
Re
sea
r
ch
Comm
on P
r
o
g
ram
of Beiji
ng
Munici
pal Co
mmission of
Educatio
n (No. KM20131
0
0110
09).
Referen
ces
[1]
F
P
Brooks.
T
he M
y
thic
al Ma
n
-Month.
Boston
: Addison-W
e
s
l
e
y
. 1
975: 2-
15.
[2]
FW McFarlan.
Portfolio appr
oach to information s
y
stems.
H
a
rva
r
d Bu
si
ne
ss Re
v
. 1
981
; 59(5):
14
2
–
150.
[3]
J Jiang, G Klein. Soft
w
a
re
deve
l
opm
ent risks to project
effectiveness.
Journa
l of Systems a
n
d
Software
. 2000
; 52(1): 3–1
0.
[4]
JJ Jiang, G Klein, R
Discenz
a. Information
s
y
st
em succ
es
s as impacted by
r
i
sks and
developm
ent
strategies.
IEEE Transactions
on E
ngineering
Management
. 2001; 48(
1): 46–5
5.
[5]
PR Garve
y
, DJ
Phair, JA W
i
ls
on. An Inf
o
rma
tion Arch
itectur
e
for R
i
sk Ass
e
ssment a
n
d
Mana
geme
n
t.
IEEE Software
. 1997; 1
4
(3): 2
5–3
4.
[6]
RL Di
llo
n, ME Pate-Cor
nel
l, SD Guikema. O
p
timal
Us
e of B
udg
et Reserv
e
s
to Minimiz
e
T
e
chn
i
cal
an
d
Management Failur
e
Risks During
Complex Project
Dev
e
lopment.
IEEE Transactions on Engineeri
n
g
Mana
ge
me
nt
. 200
5; 52(3): 38
2–3
95.
[7]
JE Ramirez-M
a
rquez, BJ Sauser
. S
y
stem
Devel
opm
ent
Plan
nin
g
via S
y
stem Matur
i
t
y
Optimization
.
IEEE Transactions on En
gineering
Managem
e
nt
. 2009; 5
6
(
3): 533–
54
8.
[8]
H Kham
oos
hi,
DF
Cioffi. P
r
ogram
Risk
Conti
nge
nc
y
Budg
et Pla
n
n
i
ng.
IEEE Transactions on
Engi
neer
in
g Mana
ge
me
nt
. 20
09; 56(1): 1
71–
179.
[9]
U Ojiako, M Ashlei
gh, M Chip
ulu, S Magu
ire.
Le
arnin
g
an
d tea
c
hin
g
chal
le
n
ges in pr
ojec
t
mana
geme
n
t.
Internati
o
n
a
l Jo
urna
l of Project
Manag
e
m
e
n
t
. 201
1; 29(3): 26
8–2
78.
[10]
A
T
ouran. Cal
c
ulati
on of co
nt
ing
enc
y in c
onstructio
n
pr
ojects.
IEEE Transactions on Engineering
Mana
ge
me
nt
. 200
3; 50(2): 13
5–1
40.
[11]
M Ran
a
sin
g
h
e
. Conti
n
g
enc
y a
l
l
o
catio
n
and m
a
n
age
ment for bu
il
din
g
pro
j
ects.
Constructi
o
n
Mana
ge
me
nt a
nd Econ
o
m
ics
.
1994; 1
2
(3): 2
33–
24
3.
[12]
KT Yeo. Risks, Classificati
on
of
Estimates, and Co
nting
enc
y
Man
a
g
e
ment
.
Journal of Ma
nag
e
m
ent i
n
Engi
neer
in
g
. 1990; 6(4): 4
58–
470.
[13]
T
W
K
w
an, HK
N Le
ung. A R
i
sk Mana
gem
e
n
t
Method
ol
og
y for Pro
j
ect
Risk De
pe
nde
ncies.
IE
EE
T
r
ansactio
n
s o
n
Softw
are Engin
eeri
n
g
. 20
1
1
; 37(5): 63
5–6
48.
[14]
J René van D
o
rp.
A Distributi
on for Model
in
g Dep
end
enc
e
Cause
d
by Co
mmo
n
Risk F
a
ctors
. ESREL
Confer
ence Pr
ocee
din
g
s, Net
herl
ands. 2
003
: 551–5
58.
[15]
MR Duffe
y, J
R
Van
D
o
rp. R
i
s
k
an
al
ysis
for
l
a
rge
en
gi
neer
i
ng
proj
ects: M
ode
lin
g c
o
st u
n
certai
nt
y for
ship pr
oducti
on
activities.
Jour
nal of Eng
i
n
eer
ing Va
luati
on a
nd Cost Ana
l
ys
is
. 1998; 2: 28
5–3
01
[16]
JR.van Dor
p
. Statistical de
p
end
enc
e throu
gh commo
n ri
sk factors: W
i
th app
lic
ations
in unc
ertai
n
t
y
analy
sis.
Euro
pea
n Jour
nal o
f
Operationa
l R
e
searc
h
.
200
5; 161(1): 24
0–
2
55.
[17]
JR van
Dor
p
, MR Duffe
y. St
atistical
de
pen
denc
e i
n
risk
ana
l
y
sis for
pr
oject n
e
t
w
orks
usin
g Mo
nt
e
Carlo met
hods.
Internation
a
l J
ourn
a
l of Prod
uction Eco
n
o
m
ics
. 1999; 5
8
(1
): 17–29.
[18]
C Genest, J Mackay
.
T
he Joy
of C
o
p
u
la
s: Bivariate
Di
stributio
ns
w
i
t
h
Uniform Marginals. T
he
America
n
Statistician.
19
86; 4
0
(4): 280-
28
3.
[19]
RB Nels
en. An
Introduction to
Copu
las. Spri
nger. Ne
w
Y
o
r
k
. 2006: 22
6-2
32.
[20]
DF
Cioffi, H Khamo
o
shi. A p
r
actical meth
o
d
of determin
i
n
g
proj
ect
risk contin
ge
nc
y
bu
dgets.
Journ
a
l
of the Operatio
nal R
e
searc
h
Society
. 200
9; 60(4): 56
5–
571
.
[21]
S Chatterje
e
, RM W
i
seman,
A F
i
egen
bau
m, CE
Devers
. Integrating B
ehav
iour
al a
n
d
Econom
ic
Conc
epts of Risk into Strate
gic Man
agem
e
n
t: the
T
w
ai
n Shal
l Meet.
Lo
ng Ra
nge Pl
a
nni
ng
. 20
03;
36(1): 61
–7
9.
[22]
T
Williams. T
h
e t
w
o
-
dim
ens
io
nalit
y
of proj
ect risk.
Internatio
nal J
ourn
a
l of
Project Ma
nag
ement
. 1
996;
14(3): 18
5–
186
.
[23]
D Wu, H Song, M Li, C Cai, J Li.
Mod
e
ling r
i
sk factor
s dep
end
enc
e
usin
g Co
pul
a
meth
od for
assessi
ng softw
are schedu
le
risk
. 2nd Internatio
nal C
onfe
r
ence o
n
Softw
a
r
e En
gin
eer
i
ng an
d Da
t
a
Minin
g
SEDM
201
0. Che
ngd
u. 2010: 5
71–
5
74.
[24]
PW
Holl
and,
PR Ros
e
n
bau
m. Cond
itio
nal
Asso
ciati
on
a
nd U
n
i
d
ime
n
si
ona
lit
y i
n
Mo
n
o
tone
Lat
en
t
Variable Models.
T
he Annals
of Statistics
. 1986; 14(4): 1
523
–15
43.
[25]
DJ Bartholome
w
, M Knott, I M
oustaki. Latent Variable Models
and Factor Analysis: A Unified
Appro
a
ch. W
e
st Sussex: Joh
n
W
ile
y
an
d So
ns. 2011: 5
8
-5
9.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A New Statist
i
cal Mod
e
l to Estim
a
te Info
rm
ation Syste
m
Contingen
cy Bud
get (Ju
n
wei Ze
ng
)
4623
[26]
RM Cook
e, R.\ W
a
ij, Monte C
a
rlo S
a
mpl
i
ng f
o
r
Gener
aliz
ed
Kno
w
l
e
d
ge
D
epe
nd
ence
w
i
t
h
App
licati
o
n
to Human R
e
li
abil
i
t
y
.
Risk A
n
alysis
. 19
86; 6(
3): 335–
34
3.
[27]
V Musoli
no, A
Pievato
l
o, E T
i
roni
, A statistic
a
l ap
pro
a
ch to
electric
al stora
ge sizi
ng
w
i
t
h
app
licati
on to
the recover
y
of
brakin
g en
erg
y
.
Ener
gy
. 201
1; 36(11): 6
697
–67
04.
[28]
CN H
aas. On
mode
lin
g c
o
rre
lated
ran
dom
varia
b
les
in
ris
k
assessm
ent.
Risk
an
alysis
.
19
99;
19(6)
:
120
5–
14.
[29]
F
Durant
e, C
Sempi. C
o
p
u
la
T
heor
y
:
an I
n
troducti
on.
C
opu
la T
h
eory
and
Its Appl
ic
ations
. 20
10
;
198(
1): 3–3
1.
Evaluation Warning : The document was created with Spire.PDF for Python.