Escolar Documentos
Profissional Documentos
Cultura Documentos
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
M.E. Sharpe, Inc. is collaborating with JSTOR to digitize, preserve and extend access to Journal of
Management Information Systems.
http://www.jstor.org
and System
AntecedentsofInformation
Quality:An EmpiricalExamination
WithintheContextofData Warehousing
R. RYAN NELSON, PETER A. TODD, AND BARBARAH. WIXOM
of
andtheDirectoroftheCenterfortheManagement
R. RyanNelson is a Professor
of
at
the
of
Commerce
the
Mclntire
School
at
Information
University
Technology
are in theareasof technology
adoptionand innovaVirginia.His researchinterests
His research
humancapital,andprojectmanagement.
tion,information
technology
hasbeenpublishedin suchjournalsas JournalofManagement
Systems,
Information
and Communications
MIS Quarterly,
oftheACM.
andAssociateDean for
Peter A. Todd is theChesapeakeand PotomacProfessor
of VirGraduateProgramsat theMclntireSchool of Commerceat theUniversity
and
of
areas
are
in
the
interests
His
research
innovation,
technology
adoption
ginia.
He haspublished
interaction.
andhuman-computer
theroleofIT indecision-making
Research.
andInformation
MIS Quarterly
ofjournalsincluding
in a variety
Systems
servesas a senioreditorforMIS Quarterly.
He currently
attheMclntire
Schoolof
ofCommerce
Barbara H. Wixomis anAssociateProfessor
of Virginia.She is a Fellow of theData Warehousing
Commerceat theUniversity
AssociateEditorfortheJournalofBusiforherdatawarehousing
Institute
research,
winneroftheSocietyforInformation
anda two-time
nessIntelligence,
Management
classesin
Wixomteachesgraduateandundergraduate
Professor
PaperCompetition.
focusesondata
andherresearch
andIT strategy,
datamanagement,
datawarehousing,
anduse. She has publishedinjournalsthatincludeJournalof
benefits
warehousing
Research.
andInformation
MIS Quarterly,
Systems
Systems,
Information
Management
is largely
ofinformation
thesuccessful
Abstract: Understanding
technology
adoption
and usage.Althelinkagesamongquality,satisfaction,
based uponunderstanding
havebeenwell studiedin theinformaand usageconstructs
thoughthesatisfaction
andsystem
toinformation
limited
attention
has
been
there
tionsystems
literature,
only
we
this
To
address
the
decade.
over
shortcoming, developeda model
past
quality
of
determinants
nine
fundamental
of
technology
qualityinan information
consisting
fourundertherubricof information
context,
quality(theoutputof an information
reprocessingsystem
system)andfivethatdescribesystemquality(theinformation
model
of
our
the
examined
We
then
the
to
aptness
empirically
quired produce output).
that
organizations
usinga sampleof465 datawarehouseusersfromsevendifferent
tools.The
andanalyticalbusinessintelligence
query-based,
employedreport-based,
and
ofoverallinformation
areindeedpredictive
resultssuggestthatourdeterminants
balance
strikes
a
model
that
our
and
in
warehouse
data
environments,
system
quality
We concludewitha discussionof the
and parsimony.
betweencomprehensiveness
ofinformation
andimplementation
andthedevelopment
forboththeory
implications
in practice.
applications
technology
/Spring2005,Vol.21, No. 4, pp. 199-235.
JournalofManagement
Systems
Information
2005 M.E. Sharpe,Inc.
0742-1222/2005 $9.50 + 0.00.
200
The QualityConstruct
There are multiple perspectiveson qualityin thebusinessliterature.
In a comReeves
and
Bednar
four
dominant
viewsofquality:
review,
prehensive
[65] identify
20 1
withspecifications,
qualityas excellence,qualityas value,qualityas conformance
The excellenceview suggeststhatqualityis
and qualityas meetingexpectations.
Thevalueperspective
thatnotiontosugrefines
assessedon someabsolutestandard.
tothecostsofachieving
ofexcellenceneedtobe assessedrelative
gestthatthestandards
ideas
to suggestthatquality
viewfurther
these
them.Theconformance
systematizes
and quantifiable
be assessedin termsof a consistent
deliveryof valuerelativeto a
expectations
suggests
specificdesignideal.Finally,thenotionofqualityas meeting
thatmayrelateto
to customer
thatqualityis definedby conformance
expectations
in shapingtheir
thataresalientto consumers
excellence,value,andotherattributes
ofquality.
perceptions
relativetoexpectations
ReevesandBednar[65] notethatqualityassessments
repon quality,withthecriticalexemplarbeing
resentthemostpervasiveperspective
servicequality.Zeithamlet al. [88] defineservicequalityas thedegreeto whicha
a set of
serviceexceedscustomerexpectations.
Further,
theyempirically
identify
aboutservice
customerexpectations
thatcollectivelydetermine
serviceattributes
includeresponsiveness,
assurance,tangireliability,
quality;theseserviceattributes
salient
beliefs
about
notion
that
with
the
Consistent
andempathy.
objectsand
bility,
havebeenempirically
tested
behaviors
[26],thesefivefactors
shapebroaderattitudes
in shapingservicequality.
toestablishtheiroverallutility
ofsettings
acrossa variety
In theinformation
referenced,
(IS) literature,
qualityhasbeena frequently
systems
withthe
construct
butrelatively
ill-defined,
(e.g., [4, 8, 41, 51, 71]). Furthermore,
ofIT servicequality(e.g.,[37,40, 63]),thestudyofqualityas a keydepenexception
Thefollowing
dentvariablehasbeenlargelysupplanted
byusageintheIS literature.
conwhiledevelopingthetheoretical
literature
sectionprovidesa reviewofrelevant
textforthequalityofIS.
ContextforQuality
TheTheoretical
have beencreatedto place qualityintoa broadertheoretical
Some IT frameworks
fromShannonandWeaver[69]andMason[52],DeLone
context.
Buildingonconcepts
andsystemqualityas thekeyinitialantecedinformation
andMcLean [22] identify
thesenotions,Seddon[67] developeda respecified
entsforIS success.Extending
modelofIS success,whichshowsthatinformation
qualityjointly
qualityandsystem
measuresof systembenefit,
influence
by perceivedusefulrepresented
perceptual
withuse).These,
nessandusersatisfaction
(whichSeddon[67] definesas satisfaction
aboutthebenefitsof futureuse, and subsequently,
in turn,influenceexpectations
actualusage of IT, whichcan have a seriesof positiveor negativeorganizational
consequences[46, 50].
and
havefocusedon empirically
Relatedefforts
assessingtheroleofinformation
in
a
of
and
satisfaction
of
as
antecedents
[64,
variety settings 68,
usage
quality
system
level.However,itis clear
at
a
holistic
treat
studies
such
70,84,85]. Ingeneral,
quality
Goodhue
are multidimensional
thatqualityconstructs
[42, 64, 67]. Moregenerally,
basis
for
the
determinacan
a
of
IS
attributes
user
evaluations
that
notes
provide
[31]
tionofIS value.
202
Information
Quality
forinformation
a variety
ofdefinitions
Researchers
haveintroduced
(ordata)quality.
viewofinformation
In general,thedefinitions
takeeitheran intrinsic
ora contextual
ofinformation
Theintrinsic
viewconsiderstheproperties
largelyinisolation
quality.
a measureof
viewreflects
froma specificuser,task,orapplication.
Thus,theintrinsic
values
thedata
an
the
actual
the
IS
and
between
data
values
agreement
by
presented
in therealworld[47,60], thedegreeto whichdatavaluesarenotinaccurepresents
andinconsistent
rate,outdated,
[48],andtheaccuracyofinformation
byan
generated
limited
IS [31, 67, 82]. Althoughthisis an important
it
is
somewhat
perspective,
because it treatsinformation
as an objectthatcan be assessedin isolationof the
towhichitis applied.Thus,intrinsic
context
butnotsufficient,
qualityis a necessary,
condition
to determine
information
quality.
A context-based
viewextendsthenotionofinformation
thatit
quality,suggesting
needstobe definedrelativetotheuseroftheinformation,
thetaskbeingcompleted,
andtheapplication
information
[47,60]. Fromthisperspective,
beingemployed
quality
is assessedbythedegreetowhichitis helpfulincompleting
a particular
task[27,45,
in termsof the
62, 72, 74, 81, 82]. For example,thismightbe assessedabstractly
usefulness
oftheinformation
in aidingdecision-making.
The contextviewexpands
thedimensions
of information
such
qualitybeyondaccuracyto includedimensions
as relevance,
andcurrency
oftheinformation
thatshapeperceptions
of
completeness,
in
the
context
of
use
[82].
quality
In additiontointrinsic
andcontext-based
dimensions
ofinformation
Wang
quality,
and Strong[82] also suggestthatthereis a representational
dimension.
The roleof
ininformation
format
anddecision-making
haslongbeena topicofstudy
processing
inIS research(e.g.,[11,36,76, 79, 83]). The representational
dimension
the
reflects
to
which
information
facilitates
and
degree
presentation
effectively
interpretation
theformat
oftheinformation
is an important
of
dimension
therefore,
understanding;
information
quality[64].
therearemyriad
dimensions
thatcanbe considered
underthelabelof
Collectively,
andrepresentational
information
andthereis littleconintrinsic,
contextual,
quality,
203
a completeandyetparsimonious
sensuson whatconstitutes
setofinformation
qualdimensions
on
the
of
andrepcontextual,
[80]. Building
categorization intrinsic,
ity
dimensions
resentational
providedbyWangandStrong[82],we havedistilleda core
as follows:accuracy(reflecting
setofinformation
intrinsic
qualitydimensions
qualcontextual
and
and
format
(reflect(reflecting
quality),
ity),completeness currency
theirderivation,
andtreatment
inprior
quality).Thedimensions,
ingrepresentational
areshowninAppendixA.
literature
definedas thecorrectness
in themappingof stored
Accuracyis mostcommonly
statein therealworldthattheinformation
information
to theappropriate
represents
refinethenotionofaccuracyto includethe
[5, 27, 43]. WandandWang[80] further
notonlyis correct,
and objective,butalso
idea thattheinformation
unambiguous,
The
of
this
refinement
is thenotionthatthere
believable.
element
and
key
meaningful
to
Information
notonlymustbe acis an important
component accuracy.
perceptual
to thenoextension
curatebutmustalso be perceivedto be accurate[82].A further
to thecorrectness
of the
tionof accuracyis consistency
[5, 27, 34, 41], referring
of
and
of
items
information
information
over
or
between
amongmultiple
relationship
of correcttime.In judgingaccuracy,we wouldassertthatusersassessperceptions
fromsystemsovera protracted
extracted
nessof information
periodof time.Their
oftheinforcorrectness
overallsenseofaccuracymaybe shapedbytheunderlying
the
of
of
the
and
of
the
information,
mation,perceptions
believability
consistency
longitudinal
experiences.
also can be shapedbycompleteness.
Beyondaccuracy,thequalityof information
which
all
to
refers
to
the
possiblestatesrelevantto theuser
degree
Completeness
to
in
information
the
stored
arerepresented
[5,27, 34, 80]. Itis important
population
ofcompleteness
onlycan be maderelativetotheconrecognizethattheassessment
textualdemandsoftheuserandthatthesystemmaybe completeas faras one useris
Whilecompleteness
is a design
in theeyes of another.
butincomplete
concerned,
and
ofthe
the
based
on
collective
is
its
assessment
experience perceptions
objective,
systemusers.
as an important
factorin
has beenidentified
In additiontocompleteness,
currency
refers
to
the
information
contextual
degreeto
quality[4, 8, 16, 35, 54]. Currency
which
the
information
the
to
is up to date,or
whichinformation
precisely
degree
is a contextual
reflectsthecurrentstateof theworldthatit represents.2
Currency
is dependent
on taskand
of systemqualityto theextentthatitsassessment
attribute
for
demands
have
different
Users
userperceptions
and,as a conse[6].
currency
may
foronetaskmaybe viewedas toodated
thatis viewedas current
quence,information
relativeto thetaskdemandsover
of currency
foranother.
Again,userperceptions
ofinformation
determinant
timewillbe an important
quality.
inTable 1 isformat.Formatis
ofinformation
Thefinaldimension
qualitycaptured
tiedto thenotionof representational
quality[4, 47, 53, 82]. Formatrefersto the
in a mannerthatis understandable
andinis
which
information
to
presented
degree
There
is
in
the
of
a
task.
and
thus
aids
the
to
user,
significant
completion
terpretable
conclusionfromthis
and theone consistent
researchon information
presentation,
is highlycontingent
ofa particular
lineofresearchis thatthesuitability
presentation
204
Table 1. Information
QualityDimensions
Dimension
Definition
Accuracy
Completeness
Currency
Format
represents.
The degree to which informationis presented in
a manner that is understandable and
interpretableto the user and thus aids in the
completion of a task.
Information
qualitycategory
Intrinsic
Extrinsic;
contextual
Extrinsic;
representational
205
is theextenttowhichthisexplanation
Whatremainstobe determined
is completein
in a globalassessment
thesenseof explainingthevariation
of information
quality.
theempiricalstudythatwillprovidethisassessment,
we turnour
Priortodescribing
ofthedimensions
ofsystemquality.
to an assessment
attention
SystemQuality
treatment
thaninformation
Overall,system
qualityin
qualityhasreceivedlessformal
In addition,
elementsofsystemqualityoftenareintermingled
with
theIS literature.
thatarecloselyrelatedto servicequalityandease of use. Forexample,
dimensions
of systemdimensions
thatrelateto IS serBaileyand Pearson[4] includea variety
Rai et al. [64] equatesystemqualitywith
vicesin theirstudiesof usersatisfaction.
of
of
ease
use.
measures
Althoughtheseconstructs
clearlyare related,
operational
is
use
A
that
to
be
to
same.
are
not
the
easy
mayalso be perperceived
system
they
ease of use maybe a consequenceof system
ceivedto be highquality;therefore,
bundledwitha highlevelofIT servicemaybe viewedto
systems
quality.Similarly,
be ofhigherquality,
makingservicequalitya covariateto thequalityofthesystem.
of systemsuccessmodels(e.g., [67]).
arethecornerstone
Such interrelationships
in
toensureconceptual
makeitall themoreimportant
Theseinterrelationships
clarity
In thisregard,we wouldsuggestthat
ofconstructs.
anddistinction
thespecification
to systemqualitythataredistinct
thatactas antecedents
thereareuniquedimensions
factors.
fromeitherease ofuse or service
To theextentthatinformation
qualityis relatedmostcloselytotheoutputofan IS,
reflects
the
processingsystemrequiredto producethat
information
systemquality
ofinteracuserperceptions
of
dimensions
the
Thus,
systemqualityrepresent
output.
should
be pertionwiththesystemovertime.In thissense,higher-quality
systems
anduse [21].
havehigherlevelsofusefulness
ceivedas easiertouse and,ultimately,
with
context
thegoal
within
an
occurs
interaction
organizational
typically
System
of sysdimensions
it
is
useful
to
consider
a particular
ofcompleting
task;therefore,
are
thatrangesfromsystemtotask.Systemdimensions
temqualityusinga spectrum
uses and
acrossdifferent
of a systemthatare largelyinvariant
thosecharacteristics
of task,context,or application.Task dimensionsare
can be assessedindependent
willdependon thetaskandsetting.
thoseforwhichan assessment
Drawingon over
oftheliterature
of systemquality,ourassessment
20 studiesthatdefinedimensions
inTable2 anddetailedinAppendixB) suggeststhattherearefivekey
(summarized
to systemquality:accessibility,
dimensions
flexibility,
responsetime,and
reliability,
to
a
and
are,
largeextent,systemdimensions.
Accessibility reliability
integration.
thatare largelyindependent
of usage.Response
definedproperties
Theyrepresent
thatareperhapsbestevaluatedin
arecharacteristics
andintegration
time,flexibility,
task-related.
ofspecifictasksandshouldbe considered
thecontext
Below,we explore
eachin turn.
it conthedegreeto whicha systemand theinformation
Accessibility
represents
low effort
tainscanbe accessedwithrelatively
[4,53, 54, 82].Accesstoinformation
to
forsystemquality.It is a systemproperty
condition
can be viewedas a necessary
206
Table2. SystemQualityDimensions
Dimension
Definition
Accessibility
Reliability
Response time
Flexibility
Integration
Systemquality
category
System-related
Task-related
in priorliterature
are shownin Appendix
Note: The dimensions,theirderivation,and treatment
B.
207
ofquality.The relativeimportance
offlexibility
be expectedtobe a keydeterminant
on
the
to
which
task
demands
indetermining
degree
changeover
qualitymaydepend
is less
forexample,we might
time.In a datawarehouse
context,
expectthatflexibility
forstatic
in thecontextofpredefined
(whichprovideinformation
reports
important
forquerying
and analysis,whichare less structured
and
tasks)and moreimportant
time.
morelikelytochangeover
thecombinarefers
tothedegreeto whicha systemfacilitates
Finally,integration
fromvarioussourcesto supportbusinessdecisions[4, 53, 82].
tionof information
and thus,integration
will varyacrosstasksand contexts,
The needforintegration
will require
Tasksthatare moreinterdependent
a task-related
property.
represents
thatsupport
toa greater
thatfacilitate
largely
degreethansystems
integration
systems
tasks
independent [32].
andSystemQuality
ofInformation
Determinants
it is usefulto thinkof
and systemqualitytogether,
information
Whenconsidering
ofa systemandthesystemas theinformation
as theproduct
information
processing
of
[22]. As notedabove,thekeydimensions
systemthatproducestheinformation
Thekeydeterandformat.
information
currency,
qualityareaccuracy,completeness,
and
of systemqualityareaccessibility,
minants
flexibility,
reliability,
responsetime,
and
should
information
these
determinants
system
explain
Collectively,
integration.
with
aboutsatisfaction
userperceptions
shouldinfluence
andtheyindirectly
quality,
andsystem(see Figure1).
theinformation
a user's
factors
thatsystem
theliterature
As explained
earlier,
mayinfluence
suggests
the
the
information
or
satisfaction
with,
of,
by system[12].Moreprovided
perception
indifferentiating
over,pastconfusion
qualityfactors
qualityfrominformation
system
effectsmayexist
(see AppendicesA and B) suggeststhatcrossoveror interaction
theresearchmodelincludescrossoverrelabetweenthetwoconstructs.
Therefore,
and
from
(systemandinformaquality(information system)to satisfaction
tionships
andsystem
effect
ofinformation
qualityon information
tion)as wellas an interaction
and systemsatisfaction
satisfaction
(see Figure1). These relationships
explorethe
exist.
that
more
may
relationships
complexquality/satisfaction
possibility
EmpiricalStudy
A cross-sectional surveywas conducted totestthemodelinFigure1. Theconwitha datawarehouse.Specifically,
textof thesurveywas userexperiences
survey
of
with
three
business
on
their
to
were
asked
types
experiences
report
participants
toolsmostcommonly
employedto access and analyzedatawarehouse
intelligence
information:
software,
(2) querytools,and(3) analysistools.
(1) predefined
reporting
that
is
set
software
is
Predefined
projectteamand
by
up thedatawarehouse
reporting
information.
is runbyuserson a regularbasisto providepredetermined
Querytools
innonroutine
to satisfy
forthemselves
information
allowuserstoextract
unplanned,
ofinformation
andmodeling
toolsallowthemanipulation
needs.Analytical
formation
208
IcOMPLETENESsl
'
ACCURACY
V^
^
^/information]
i quality t
V^
( FORMAT
/information^
^satisfaction/*
'
'
(currency)^
^ - */
'
N,
/
' /
/infosaA
/ '
(reliability!
FLEXIBILITY
k,^^
^-
'
'
^7
X^
( ACCESSIBILITY
(RESPONSE y
I
TIME I
'SYSTEMSAy
SYSTEMY
QUALITY I
'
SYSTEMY
^SATISFACTION/
/
/
I INTEGRATION
I
ofInformation
andSystemQuality
Figure1. Determinants
209
froma datawarehouse.
Thefollowing
extracted
sectionsdescribethestudy,
including
and
theinstrument
the
development
process
samplethatwas used.
Instrument
Development
ofthesurveyinstrument
followedtheprocessproposedbyMooreand
Development
Benbasat[55].A literature
reviewwas conducted
tolocatepastoperational
measures
underinvestigation,
of
were
oftheconstructs
from
valigroups questions
compiled
to represent
each construct,
and wordingwas modifiedto fitthe
datedinstruments
and graduatestudents
data warehousecontextto be studied.Next,tenprofessors
sortedtheitemsintoseparatecategories,
identifying
ambiguousor poorlyworded
items.Itemswereremovedandminorwording
changesweremadepriorto a second
The threeitemsthatwere
whichdidnotuncoverfurther
roundofsorting,
problems.
wereselectedforeach qualitydimensionand included
mostaccurately
categorized
ina randomorder.3
Eachquestionwas measuredona sevenonthesurveyinstrument
from
scale,ranging
(1) strongly
disagreeto (7) strongly
agree.
pointLikert-type
wasfurther
theinstrument
reviewedbyacademics
thesurvey,
Beforeimplementing
anddatawarehousing.
withknowledge
ofsurveydesign,IS quality,
andpractitioners
The resulting
Minorchangesweremadebasedon theirsuggestions.
surveywas then
to
from
a
problems
largepublicuniversity identify
usingrespondents
pilot-tested
and procedures.
For thispilottest,
withtheinstrument's
content,
format,
wording,
datawareto 250 activeusersofa largepublicuniversity's
surveysweredistributed
Pilot
in
a
rate.
com29
73
house; responded,
response
participants
resulting
percent
written
comments
about
and
and provided
length,wording,
pletedtheinstruments
wereinterviewed
togaina richerunderstanding
Two oftheparticipants
instructions.
theinternal
Thedatawereanalyzedregarding
ofthefeedback.
consistency
reliability
the
0.7
and
each
exceeded
Cronbach's
oftheconstructs
accepted levelof
alpha,
using
[58].
reliability
weremadeto the
Based on theresultsof thepilotsample,minormodifications
the
constructs
fromFigure
included
items
final
The
survey
measuring
surveydesign.
and self-reported
items.
The
1 as well as a seriesof demographic
usage
specified
areshowninTable3.
itemsanddescriptive
statistics,
organizedbyconstruct,
Sample
senttomembers
ofthe
weresolicitedvia an e-mailannouncement
Studyparticipants
a freestudyto assess the success of their
Data WarehousingInstituteoffering
ofindusSevenorganizations
froma variety
datawarehousesoftware.
organization's
tries(e.g.,healthcare,consumer
agreed
goods,financialservices,andgovernment)
was askedto distribute
Each organization
toparticipate.
surveysvia inpaper-based
All surveyswereconfimailtoall oftheactiveusersofitsdatawarehouse.
teroffice
information
was
collected.
no
dential; identifying
Completedsurveyswere
personal
and returned
to theresearchers.
collectedby a contactpersonat each organization
with
an
Table
overallstudyrevaried
across
rates
(see
4),
organizations
Response
21
465
rate
of
yielding completedsurveys.
percent,
sponse
210
Si
<
f
2,
cJ> ^
g
CO
CL(5
li
H o
lo^iT)
^ ^
io
iri^^t
h-ohk^cq
o co co
corico
coiaio
co io
io
b: P !>
oqrv.rN.
^OD00li)
in co t^
li
II
II "
bTili)^
CO
^COO
^1^.0
CN
o?T
IO "t IO
O>
O)
O)
II "(D
oo co
0)1^1^
M-in^
If)
CO CD
I
n "05
O)
ccoco
li)
li
il "
Im
T3
if) (O
I
il "
OJ
T-r^Tt
h-CDOO
il a)
E
h "55
E
il "
E
own
"^
li
Si
S!
" "05
E
^tlOlO
H O
^ IO (D T-
9?^
lO^-^-
e
I^
O
-a
-a
S
CO
i
&
1
co
cio
?^ "5. g
CO
'-3
le!
li
J3S
IS
S 8
^S"5
I gf
o ' o
>
i=
-ri
-H
fil
f
II
1|1
11
III
Ht
1
o^co-od-o
c
.2
g
c
5
p-Q.
ii
Sfl
sM
||
*> -o '>
2E2
aaa
<
Z ? o.
g|S
11!
"
HI
HI
Pd
1^1
si
ca (e co
?<
>
2
SEE
000.
-OcdE
2S0
"> "2 i
22o
Q. 7 "E
cecero
0cod)2g:
clq.^
o,
211
8
8-
S E
^
it
g
io
feS
^
EBS
*'-^
S^
il
SS
lo
loin
-*0
?n
d>
fi !
i t
li
;s
$
|t
1
5
i
&
5 S
fi
I
ooe
--
OO2-EO_==
^
o jl
H d>
II <B
TtLo^r
^tco^
S
i
ti
"O
il,
us
cr
>*
OC
I
!
III
2
S
p
cocco
^m
8.
i I 1
5 I 5
..i
HE
==
ji
C3)
----.Eoo
SEE
'
iifMi II
f ili
1 6
il 1
t lli
1 llf &1 111
o
c
~
^*C
io
"*
Po
?Bg
io in io
^^
"
^555
lo^t^t
io ^
II "a
|| -qj
"irjioio
| "-
dl
!2 S
SKR
S0
M=
TfTf-'t
F2gS
f5
S?
SSS
gg
id lo
S2 S I
*" *" 1
PS2 55
il "05
il "o
i!
S
-
in^t^
SSS
PIS
Si
Sfefe feS
8 S
io
ii
ii
II "0
II
8 fe
8"^
888
S*""
SS
8"
S5
cw
FO3
"co^3
(OCtf)
li
!i
i ||
i fit
lilil
Illc
I III | M |l|i||
.2o
oQ.Q->,
Hg
LL
ajQ-coto
OQ.
212
|^
S
2
ggg
SI*""
Sqio
I
3
SRS
S!^^^
SSP!
fe^^^
iSS
S^^r-
Sd
Si
iS
Ip^5
iriio^t
^' ^
T^T^T^
|{552
ifiioio
^i
io io
T-
1-
T-
to
OqJ.
E "g 3
o o5 > i2
^t^
SS
S^^
ojl
E
?S
^'^t
T- T-
oji
|55!fe
Ttio^t
|5S'*^
Jl
u <D
11 (D
|So
^ 10
^5^
11"
u "
l
E
S
dl
dl
o
c
lit
-^cfc
^S-cr
O)
f.
"O ^
^^ >
.S2
*"i
Ut m J V
II If! II! il
i :Pli
III
Ili
II
His
si i. g
$s$l
o>0
T3
<D
.S
en
dl
a E
^Sf2
T^l-^T^
S58
>8
^- ^i ^
in^-^
di
a E
2S
O-r
u o
E
u "
0 E
fef3
go
50
coo
11">
0 E
11
0 E
g
<
SS
8$ fe
"
8^ "
8S
g"*"
I
S
g
<
S. *
=
en
B^B=A"
iOOgi<Oyco
co
2 13
Table4. Participating
Firms
Company
Surveys
sent*
Surveys
returned
Responserate
(percent)
A. Health care
129
40
31
B. Packaged goods
300
92
31
C. Financial services
179
23
13
D. Health care
42
108
39
E. Public sector
172
14
1,200
F. Public sector
231
61
26
G. Public sector
66
35
53
Overall
21
465
2,213
* Numberof
surveyssentto each company.We cannotbe certainthatall surveyssentwere
to data warehouseusers.Thus, our effectiveresponserateis likelysomewhathigher
distributed
thanreportedhere.
Results
The research model was tested usingpartialleastsquares(PLS), a structural
models[7,
modelingtechniquethatis well suitedforassessingcomplexpredictive
of thescales used to
teststhepsychometric
18, 86]. PLS concurrently
properties
measurethevariablesin themodel(i.e., themeasurement
model)and analyzesthe
of therelationships
and direction
amongthevariables(i.e., thestructural
strength
used fortheanalysis,and thebootversion
2.91
was
[19]
model)[49]. PLS Graph
method(100 resamples)used to determine
thesignificance
of the
strapresampling
model.
within
the
structural
paths
214
Table5. StudyParticipants
Number
Organizational level
Senior management
Middle management
First-levelsupervisor
Analyst
Clerical
13
95
48
257
27
3
22
11
58
6
22
79
22
37
82
96
116
5
17
5
8
18
21
25
180
270
40
60
Functional area
Accounting
Finance
Human resources
Informationsystems
Marketingand sales
R&D
Other
Gender
Male
Female
Percent
Model
Measurement
of internal
modelincludedtheestimation
The testof themeasurement
consistency
itemsforeach ofthe
anddiscriminant
andtheconvergent
validityoftheinstrument
threetechnologies
querytools,andanalysistools).Table3 lists
(predefined
reports,
forresponsesacrosseach
reliabilities
thesurveyscalesandtheirinternal
consistency
measureswerewellabovetherecommended
All reliability
ofthethreetechnologies.
levelof0.70,indicating
[58].
consistency
adequateinternal
and discriminant
These itemsalso demonstrated
validity.
convergent
satisfactory
variance
extracted
have
an
when
constructs
is
average
Convergent
validity adequate
theAVE fromthe
discriminant
(AVE) of at least0.5 [28]. For satisfactory
validity,
andother
construct
shouldbe greater
thanthevariancesharedbetweentheconstruct
constructs
in themodel[18]. Thiswas trueforeachofourconstructs
(see Appendix
whenitemsload above 0.50 on their
C). Convergent
validityis also demonstrated
associatedfactors;
all ofthemeasureshavesignificant
loadingsabovethesuggested
threshold
(see AppendixD).
Structural
Model
The pathcoefficients
andexplainedvarianceforthestructural
modelsareshownin
of information
Figure3. Overall,thefourdeterminants
qualityexplainedover75
2 15
Tool Usage
Figure2. SampleBusinessIntelligence
216
I COMPLETEI
*^%
^'
ACCURACY
K^-*)^
^
/*
-. HflNFORMATION^
[21.426*** ^/INFORMATION!
QUALITYA P] 225
X3- **>A
^SATISFACTION^
"X
FORMAT
V^/mJ*
/ ^/
V_y
^vW
PT
'
P1-782
'
P1.761
'
J5
currency
y
'
^ - ^
S
(
^
^'
Jr*1
VW
ACCESSIBILITY
'
/infosatN
/ '
'^**!>A
/
SYSTEM Y
QUALITY
)
/[Jj31
^L*
' /
reliabilityI
'
/
bisk P1558
^^
[31.514
[21.578*** Jy SYSTEM V
Pl-WI*- ^^SATISFACTION/
&W* / ^'743
// $Jf
I w-750
[1^536
PI'486
Pi-511
/responseX/^^
I
TIME I
^sQ*
(integrationV
Figure3. ResearchModelResults.
Notes: 1 = Predefined
2 = querytools;3 = analysistools;* p < 0.05,
software;
reporting
** p < 0.01, ***/?< 0.001.
217
tools,whereasresponse
systemqualityforthethreekindsof businessintelligence
in thecase ofanalysistools.
timeis onlysignificant
influenceon systemqualityacrossthe
Reliabilityappearsto have thestrongest
tools.On average,accessibility
threebusinessintelligence
andthenflexibility
arethe
the
two
are
for
nextmostinfluential
reversed
factors,
although
positions
querytools.
to
have
a
level
while
of
effect
on
theassessThus,
reliability
appears
universally
high
weakeffect,
mentof systemquality,andintegration
,
appearsto havea consistently
ofaccessibility
andflexibility
therelativeeffect
appearstobe moretool-dependent.
we did notfindan interaction
Withinthisspecificcontextof data warehousing,
the
frominforsatisfaction.
betweensystemandinformation
And, crossovereffects
andsystem
satisfaction
werenotsignificant
mationandsystem
qualitytoinformation
it shouldbe
and
tools.
withinthecontextsof predefined
However,
reporting query
thecontext
ofanalysistools,thepathleadingfromsystem
notedthatwithin
qualityto
was significant
satisfaction
information
(0.413),andthepathfrominformation
qualThiswillbe exploredfurther
next.
was notsignificant.
satisfaction
itytoinformation
Discussion
This study develops and tests a model thatexplainshow varioussystemand
user
andsystemqualityand,ultimately,
information
influence
attributes
information
of
determinants
is
Based on empiricalresults,it appearsthatour set
satisfaction.
andsystemqualityin datawarehouseenviofoverallinformation
indeedpredictive
and
and thatourmodelstrikesa balancebetweenbeingcomprehensive
ronments
that
the
determifrom
the
results
it
is
clear
same
At
the
time,
empirical
parsimonious.
nantsofqualityarenotall equivalentin theirpredictive
powerwithinthecontextof
datawarehousing.
acrossall threedata
determinant
Forinformation
accuracyis thedominant
quality,
thisis not
attribute
of
information
core
intrinsic
As
the
warehouse
quality,
technologies.
At thesametime,thissuggests
context.
especiallyin a datawarehousing
surprising,
overallperceptions
in
thatreflects
variance
thatthereis significant
accuracy
perceived
information
of
the
it
reinforces
ofinformation
importance managing
quality.Further,
would
be
that
users
It
should
not
assumed
of
accuracyas a keydeterminantquality.
data.
as providing
initiatives
high-quality
recognizedatawarehouse
universally
thethree
across
most
influential
determinants
are
the
next
format
and
Completeness
accentuated
of
be
The
datawarehousing
may
technologies. importance completeness
ofdisparateinformation
becausetheintegration
inthedatawarehouseenvironment,
sourcesis a keydatawarehouseimplementation
objective.Formatalso has a consiswitha warehouse's
whichis againconsistent
acrossthethreetechnologies,
tenteffect
orientation.
Ourresultssuggestthatthesethreedianddecision-making
information
ofthevarianceininformation
accountforthree-quarters
mensions
qualcollectively
the
and
to
test
would
be
research
Whereas
consistency
required
subsequent
ity.
itprovidesan initialindication
thata
acrossothercontexts,
ofthisfinding
robustness
information
andmanagedtoinfluence
canbe identified
offactors
smallnumber
quality.
2 18
2 19
Limitations
beforediscussingimplications
needtobe considered
threelimitations
Thefollowing
withinthespecificdowas
conducted
research
this
forresearchandpractice.First,
can
ornotthefindings
whether
As a result,itis uncertain
mainofdatawarehousing.
Second,thisrebe appliedmorebroadlyor to otherspecificformsof technology.
andthestudycontainsthetypicallimisearchwas basedon a cross-sectional
survey,
included
associatedwiththiskindofresearch
tations
Third,thisresearch
methodology.
onlyspecificfactorsbased on priorresearch,butdid nottesttheuniversalset of
andsystemquality.
forinformation
antecedents
forResearch
Directions
andconceptuallinkagesrelated
attributes
previousstudieshaveidentified
Although
andsystem
ofinformation
focusonthedimensions
a comprehensive
toquality,
qualin
asof
the
One
IT
literature.
from
the
absent
been
has
challenges
surprisingly
ity
can
that
enhance
measures
and
definitions
find
to
is
understanding, be
sessingquality
220
andhavethepotential
toguidemanagement
action[65]. In
practically
implemented,
thisregard,
oneofthekeycontributions
ofourworkis theidentification
ofa comprehensivesetofdeterminants
thatpredicts
is relatively
andmost
quality,
parsimonious,
enhancesunderstanding.
We feelthatthismodelcan serveas a powerful
important,
lensbothin interpreting
theresultsof priorinvestigations
and in shapingrigorous
research
modelsforfuture
that
tie
to
satisfacinquiry attempt together
system
quality,
tion,anduse.
Theantecedents
ofIS usageandvaluenowcanbe evaluatedusingthelensofqualin
the
of
Morespecifically,
information
ity
design an IT artifact.
quality(completeand
and
ness, accuracy,format, currency) systemquality(reliability,
flexibility,
andresponsetime)nowcan be connected
to usersatisfacintegration,
accessibility,
tion(information
andsystemsatisfaction,
as depictedinFigure1. Our
respectively),
resultssuggesta wayto integrate
elementsof thetechnology
artifact
withtheuser
and
use
that
to
avoid
the
associated
withtakinga
perceptions
helps
potential
pitfalls
"proxyview"ofIT [59].
Further
ofthemodelshouldexamineitsrobustness
and stability
acrossIT
testing
environments.
We believethatevolvingbusinessor technological
factorsmayalter
thenatureandrelativestrength
oftherelationships
ina conceptual
model,suchas the
one developedhere.Ourresultssuggestthatsomefactorsare moreimportant
than
othersinthedatawarehousing
thatwe examined.Itis notcleariftheseresults
context
will be stableacrosstechnologies
or applications.
In fact,one wouldsupposethat
about
the
effect
relative
ofthequalityattributes
fordifferent
forms
oftechtheorizing
avenueofresearch
topursue.Contemporary
nologywouldbe an important
examples
includeWeb-basedapplications
services[17]. In additionto
[42] andmobileInternet
onlinetrading
andhomelandsecurity,
thatshouldpresent
contemporary
applications
a different
testofthemodelwouldbe reservation
andsomereal-time
invensystems
andresponsetime).
torysystems(requiring
currency,
integration,
Therealso is theopportunity
tolookatchangesinthemodelacrossdifferent
kinds
of technologies
withinthesamecontext(similarto thisstudy)withthepurposeof
characteristics
influence
thestrength
of the
investigating
exactlywhichtechnology
Forexample,thisstudyseemstosuggestthattherestrictiveness
qualityantecedents.
ofa toolmakesa difference
in shapingqualityperceptions.
othernontechniFurther,
cal characteristics,
suchas tasktypeoruserdemographics,
roles
mayplayimportant
in understanding
userswithtacticalversusstratequality.Withindatawarehousing,
access to dataversusdepartmental
access to
gic tasksor ones withenterprise-level
datamaycause different
variations
oftheresearchmodelevenwhenthesamebusinessintelligence
toolis used.Thisneedstobe studiedfurther.
In additionto examiningthestability
of resultsacrosstechnologies
and applications,themodelshouldbe testedwithalternative
on
measurement.
We
perspectives
havetakentheapproachofaskingforabsoluteassessments
ofqualityattributes
that
areinterpreted
relativetoindividual
need.Alternative
toassessingquality
approaches
based on expectation
gaps,as is typicalin theservicequalityliterature
[37], might
also be considered.
221
ofdifferent
research
couldexploretherelationship
future
In addition,
development
For
to
information
and
and
systemquality. example,do
methodologies techniques
Does extreme
more
flexible
lead
to
programapplications?
practices
object-oriented
Thus,themodelprovidesan avenuetoconcepreliability?
mingincreaseapplication
tualizehow variousdesignmethodsand techniquesmightinfluencesystemand
This providesan aptheindividualqualityattributes.
information
qualitythrough
on systemusefuleffects
to theirultimate
proachforlinkingdesignmethodologies
anduse.
ness,usability,
forPractice
Implications
ofa
topracticeinthreeimportant
Thisstudycontributes
ways.First,thedevelopment
should
to
satisfaction
and
then
to
antecedents
modellinkingspecificquality
quality
andinfluence
understand
tobetter
allowsystem
systemuse.For
designers
ultimately
studied
ourfindings
tools
business
the
three
case
of
in
the
here,
intelligence
example,
andclearly
on
be
should
that
accurate,
complete,
focusing producing
suggest designers
With
information
enhance
in
to
informationtheirquest
formatted
respectto
quality.
systemquality,designerswill have thegreatesteffecton qualityif theyfocuson
relationThecrossover
andflexiblesystem.
a reliable,accessible,integrated,
creating
thatthe
need
to
know
when
that
may
analysistools,designers
designing
shipssuggest
its
with
satisfaction
theusers'ultimate
affects
output
qualityof thesystemstrongly
As datais collectedacrosstechandfocuson improving
systemqualityaccordingly.
itshouldbecomeclearertowhatextentthere
environments,
nologiesandindifferent
and special-purpose
aredominant
factors,
quality
secondary
qualitycharacteristics,
factorsthatwillguidedesign.Thus,forexample,ifaccuracywereto emergeas the
of information
determinant
dominant
quality,itwouldsuggesttheneedto payparareembeddedin applications.
to thewayvalidationtechniques
ticularattention
of
a
can
nine
determinants
of
Second,theset
play usefulrolein thecomparison
Forexample,
off-the-shelf
fora commercial
vendorsolutions
application.
competing
solutions
businessintelligence
it wouldbe usefulto be able to comparecompeting
on bothinformation
withrespectto theirrelativeweighting
quality(completeness,
and systemquality(reliability,
and currency)
integraflexibility,
accuracy,format,
the
As we beginto understand
andresponsetime)as appropriate.
tion,accessibility,
it shouldbe possibleto use meaofthevariousqualityfactors,
relativeimportance
intheevaluationofsoftware
suresdiagnostically
options.
shouldserveas an aid toimplementation
thequalityfactors
Finally,understanding
andtrialphases,
strategies,
testing
Managerscanshapecommunication
management.
rolloutsso thattheybetterlead to positiveuserreactions.For exand technology
clarofinformation
canhelptoshapeperceptions
testing
approaches
ample,usability
to
ensure
ofinformation
qualityassuranceprocessesmaybe needed
ity.Descriptions
shouldprovidebothmanagers
various
the
to
Attention
qualitydimensions
accuracy.
toolsformanaging
additional
anddesigners
processes.
implementation
222
Conclusion
Based on the theoretical linkage betweenqualityandusage,theprimary
objectiveofthisstudywas toincreaseourunderstanding
ofinformaofthekeydimensions
tionand systemquality.More specifically,
we soughtto identify
a comprehensive,
setof determinants
thathelppredictthequalityof an IT artifact.
yetparsimonious,
Based on theliterature,
a totalofninefundamental
ofIT-related
determinants
quality
wereidentified,
fourundertherubricofinformation
quality(theoutputofan IS) and
fivethatdescribesystemquality(theinformation
toproprocessingsystem
required
ducetheoutput).
The ninedeterminants
collectively
portionofthevariance,
explaineda substantial
over75 percent
overallforinformation
and
overall
74 percent
quality approximately
forsystemquality.We believethattheseresultssupporttheaptnessof our model
withina data warehousing
context,and thatthemodelstrikesa balancebetween
and
ofhow
whileproviding
additional
comprehensiveness parsimony,
understanding
In
will
areshaped. conclusion,
we hopethatthisresearch serveas
qualityperceptions
a catalyst
foraction,encouraging
bothresearchers
tofocusonqualandpractitioners
ityas a coreconceptwithintheIT discipline.
Theauthors
thankIzakBenbasat,Dale Goodhue,StefanoGrazioli,Michael
Acknowledgments:
reviewers
fortheirhelpful
of
comments
onearlierdrafts
Morris,PeterSeddon,andanonymous
thispaper.
Notes
1. Itis estimated
that50 percent
oflargecompaniesinNorthAmericaandEuropeareusing
thebalancedscorecardformeasuring
thequalityof financial,
and human
customer,
internal,
resourcesperformance
[571.
2. Currency
is also commonlyreferred
to as timeliness
(see AppendixA) and has been
further
refined
by Bailou et al. [6] to includebothcurrency
(how
(age of data)and volatility
longan itemremainsvalid).
3. Twoquestionswereusedtoformeachofthetwoscalestomeasureinformation
satisfactionandsystemsatisfaction.
4. www.nsf.gov/od/lpa/news/03/fact030124.htm
(accessedJuly22, 2004).
References
1. Agmon,N., andAhituv,
N. Assessingdatareliability
in an IS. JournalofManagement
4, 2 (Fall 1987),34-44.
Information
Systems,
2. Ahituv,
N.Assessingdatareliability
inan information
MIS Quarterly
4,4 (1980),
system.
61-75.
3. Anderson,
C.P. Stageoftheproduct
lifecycle,businessstrategy,
and
C.R.,andZeithaml,
businessperformance.
Journal,27, 1 (1984), 5-24.
AcademyofManagement
4. Bailey,J.E.,and Pearson,S.W. Developmentof a tool formeasuringand analyzing
usersatisfaction.
Science,29, 5 (1983), 530-545.
computer
Management
5. Bailou,D.P., and Pazer,H.L. Modelingdataand processqualityin multi-input,
multiScience,31, 2 (1985), 150-162.
outputinformation
systems.
Management
223
224
225
59. Orlikowski,
W.,andIacono,C.S. Researchcommentary:
Desperately
seekingthe"IT" in
- A call totheorizing
theIT artifact.
IT research
Research,12,2 (2001),
Information
Systems
121-134.
Communications
60. Orr,K. Data qualityand systemstheory.
of theACM, 41, 2 (1998),
66-71.
and FamiliesCan Benefit
61. Perlow,L.A. FindingTime:How Corporations,
Individuals,
Practices.Ithaca:CornellUniversity
Press,1997.
fromNew Work
Communications
62. Pipino,L.L.; Lee, Y.W.;andWang,R.Y.Data qualityassessment.
ofthe
ACM,45, 4(2002), 211-218.
63. Pitt,L.; Watson,R.T.; and Kavan,C.B. Servicequality:A measureofinformation
sysMIS Quarterly,
temseffectiveness.
19, 2 (1995), 173-187.
64. Rai,A.; Lang,S.S.; and Welker,R.B. Assessingthevalidityof IS successmodels:An
Research,13, 1 (2002), 50-69.
Systems
analysis.Information
empiricaltestandtheoretical
andimplications.
65. Reeves,C, andBednar,D.A. Definingquality:Alternatives
Academy
Review,19,3 (1994), 419^45.
ofManagement
information
andmanagement
66. Robey,D. Userattitudes
systemuse.AcademyofManagementJournal,
22, 3 (1979), 527-538.
and extensionof theDeLone and McLean modelof IS
67. Seddon,P. A respecification
success.Information
Research,8, 3 (1997), 240-253.
Systems
oftheDeLone andMcLean
68. Seddon,P.,andKiew,M.-Y.A partialtestanddevelopment
modelof IS success.In J.I.DeGross,S.L. Huff,and M.C. Munro(eds.), Proceedingsofthe
Atlanta:AssociationforInformation
on Information
International
SysSystems.
Conference
tems,1994,pp. 99-110.
Urbana:
69. Shannon,C.E., andWeaver,W. TheMathematical
TheoryofCommunication.
ofIllinoisPress,1949.
University
dimensions
ofIT infrastructure
thetechnological
70. Shaw,N.G.Capturing
change:A model
andempiricalevidence.JournaloftheAIS,2, 8 (2002), 1-33.
Associationsandimplicameasuresof systemeffectiveness:
A. Alternative
71. Srinivasan,
tions.MIS Quarterly,
9, 3 (1985), 243-253.
Communications
72. Strong,
D.M.; Lee,Y.W.;andWang,R.Y. Data qualityincontext.
ofthe
ACM,40, 5 (1997), 103-110.
- Appreciation
and involvement.
73. Swanson,E. B. Managementinformation
systems
178.
2
2,
Science,
(1974),
Management
dataquality.Communications
74. Tayi,G.K.,andBailou,D.P. Examining
oftheACM,41, 2
(1998), 54-57.
information
75. Taylor,S., andTodd,P.A.Understanding
usage:A testofcomtechnology
Research,6, 2 (1995), 144-176.
petingmodels.Information
Systems
Effects
ofpresentation
orgoldgraph?
76. Tractinsky,
N., andMeyer,J.Chartjunk
objectives
MIS Quarterly,
on information
andcontent
23, 3 (1999), 397-421.
processing.
desirability
extensionof thetechnology
77. Venkatesh,
V., and Davis, F.D. A theoretical
acceptance
fieldstudies.Management
model:Fourlongitudinal
Science,46, 2 (2000), 186-204.
78. Vessey,I. Cognitivetit:A theory-based
analysisot thegraphsversustablesliterature.
DecisionSciences,22, 2 (1991), 219-240.
79. Vessey,I., andGalletta,D. Cognitivefit:An empiricalstudyofinformation
acquisition.
Research,2, 12 (1991), 63-84.
Information
Systems
Communidataqualityinontologicaldimensions.
80. Wand,Y, andWang,R.Y.Anchoring
cationsoftheACM,39, 11 (1996), 86-96.
Communications
on totaldataqualitymanagement.
8 1. Wang,R.Y.A product
of
perspective
theACM,41, 2 (1998), 58-65.
82. Wang,R.Y, andStrong,D.M. Beyondaccuracy:Whatdataqualitymeansto dataconsumers.JournalofManagement
12,4 (Spring1996),5-34.
Systems,
Information
tabularandtextualmethods
J.N.An experimental
83. Washburne,
studyofvariousgraphic,
material.
JournalofEducationalResearch,18, 6 (1927),465-476.
ofpresenting
quantitative
Towardsa theoreti84. Wixom,B.H., andTodd,P.A.Useracceptanceandusersatisfaction:
cal integration.
Research,forthcoming.
Systems
Information
ofthefactorsaffecting
data
85. Wixom,B.H., andWatson,H.J.An empiricalinvestigation
success.M/SQuarterly,
25, 1 (2001), 17-41.
warehousing
226
227
s
|
r- ,
H
U
{2
CM i00. ^.
CO
^
1
2
CO
^
.OONmQOOCOT-Nr-.^oCOOO)
CO. IT). ^. CO. CO. C'J. IO. CM. PO. CO. LO ^t. CM. .IO. CO. CO. IO. CO. CO.
li]
s e
I i
CO
il
il
<D
|
1 I
s
in
5*1
-sP
i
1
1
il
Is*
?
1<
.,
i
?
ss
I 1 MS
1
d
Hill
IH!
I
1 |5U
1 l|?
:-|53|ia
** il1-! liiifl
il iiiiiiiH
C/3
C/2
C
D
1
M
S
"3 S
H
<
Cu
<
1
U
litifiS!
iS
IJ!!l?!l!lS I PI Ili
iilllfiiliillllli
li fiillil HL
IliiffiliiiiliipiHitil^lSiSllEiEiEiEiEiEJSSSlElE^ilEiE
I
&
doo
o
oooo
i I I II ^ I _
S
&
5*^
S^
cc
&
^
&a
S4
co
cocococococococdcciO^wwcocooS^co
3
3333333333gggggfc3
<<
<
<<<<<<<<<<CQUJlIlJL
228
go- ;
5_
=5
sig
a
CO
3
!
i
!
li
Iff
fi
a s S
i i i
II
i S 11
sili
ill
s'alii
s.i&|a
I I
1 i
c
<S.O(0WS
! tlifPf
s
f
!
Q.
Si
li
f
ft
d i I
Isl ! ! 1
fi** 3
--Soi73
.81?
IlHi
11 i
E
E_-
I l
l**j-
! i li I
ifillfllll lilllpliif i
i
co
1
e
o
U
<
x
i
I
S
1 S
1
<
O
U
-5>.
^lfo|ffit|.
|.I1I1|
a
I
^(DDciDOaJIDEclB
c 2 O - DC CC CC O) 3
s i
8g8ggS8SS8S
22222i2gso
>
fffffllllii!
tlillllllRlil
OOOOOOOOOc^Q--=O<
1^
229
"
CO
If)
00
" "
p: g
S!_
IE
-^
^
o
t
*o
d>
.S
18
8 t
o
c
-
Si
r -S
Ie
ill
-S
oo
? tg
oil
llllll
| 8& I
a-Sl
1-oS
|
5
^*3S?!tfc
itili l
Unes
-o
is
o:
S"o
S I
-o
-2
"5
2
oS
i
~
c g
S.8 2j3|o
t
8 I
*^w2h
3s^
l*
-I
o^
4!ii1ftifi!f IIM
tiUiUiiiUiHim i
litliliiiisiai
CI
towwwa)
m!!!!!s
llUllll
*C
O- "2
coco
*"O
c
b i ^mii
IH
e
1co-3
^
?
.-t^c:
75"CDC/3
"'S ^ S
I l|??
I 2S$k
''S
230
C/5
CO
t
.1
If
co $
f
I
Si
^S
!- ?
i
.9-
i- ! I
i
-
co
S)
"4= J2
CO
.>
co
or
cd
+=
a) "co
"O
.& i* g I 5i S*
1 li ? i il lit
i! !
III
C/3
.2
lilfllli
2 ; 8 ? I 8 S S fi
o^0-
<
to S ^
> j(T
(B
Q. S
OF
*=
C/3
PQ
<d
*0
"
"
^:^:iix:^:^:.^.Et0jz
hHc5HHHO|-QQH
unsi
ili niiiiim
'c/3
i
!
Z
52
O
>,
S
COCOCO
COCDC
i
1
il
Q.CL
C OC
0.000)0000
CCHHHH-h-HH
231
S"
SS
I I
II
II
S
o
&
^ "
s
C
II
S'a.
li
sil
i i
li
0 E
CL
0 3
ili
Ili
I i
lit
1 Si
M
Q.
-g
it
ili ?
i% 1
%%1
ii i
il* i
f! I tIJ11i ! i
il il! Prilliill!!
JE<LUiii
E <
iiiiii
<i--ocoH^h-
null
^h
(ch
I3
c
c
S
S
8
3-
'"S
8|
5 2 1 I o
I 8 i S
<<<Oi2co
&2?
&&
ISSig
u. u.
u. u.
?5
tf
uj
il
%
-s
.
-2 i 'S
S
*
o S M
S * |<
i>
232
&
c/3
o)
Z
H- I
fe
co S SS
dddddd
oq ^
io co In 't io
ooLoincocp-^ir)
WSI
S'SSS
^^^PSrfSS
cqcoini^coi^ioir)
ddddoooo
ddddddddd
P4
'S
dddddo
dddddddd
a
o
cocDcqcpiqin
COI^^CMCVICMCD
coT-30if)CDh-cx5cx)
cqifl^ininininin
X
m
H
S S
ddddddd
4-1
ddddd
ddddd
cZh
o co co r^
dodo
SoqSr^
dodo
o o o
doo
, S?
o o
C/5
P^
(DC'|lOi-li)0)^OCVI
OJNNCDOONCMOM
qncq^inincpLqco'^;
roin^ininincpinco^;
ooddo
fj
S
n/
r^m^cocoh^ooco'ito
0)N00S0)NC0^"T-OlO
ojcoijOT-cocDCNjr^-^qj
NOOt-IOCOODt-00010
ooo
oo
oq^cDinininh-cpcqcqif)
oq^n^^iocpcpi^cDin
2
^
CJ
rj
^onioi-O'cocmnco
^<D(ONtOU)MO^O)rcqinincpiniqininNNWi)
inoocomcoococoooNCD
NWT-COlfllftOCONCMOroqincqcDco^ininNNcoin
dcicidd
dddddddddddd
pin
5
2
u
&1 ddddddddddddd
OinN^CDCMr-COOT-NO)
T-T-mNoooNcoNtooo'tin
r-^CO00CO00C0O)TfCMNO)
T-incomr-coNinoocoN^io
ooo
'-3
<
cu <LLa:Ei:?P<?a)co?
<
iex
<.cil?i-<wco
$S
233
g .1 g
21
"^
a.
O5
lo oo
<2
g >; c^
O)h-
O -ti
li
(O 00 n C
O O) CM u || O
i gl
rn o CO S P S
o S S co g|s
N. 'S 5,
m
o o o o .g g
O O
II
o * ^ <S
w co co w co
CO CO CO N N N
OOCDCOCDLOLO
ddddd
(O
O)
CM CM N
^ I 8
S^?
.2 h II
"g ^g^
-s S g
"md
O)
O)
e s
3Z3
j
evi O) m co evi
oo ^ ^ cq lo <* "* g - ^
o
ddddddo
g
.* c
o O)
00^t"^t"tOh-00
cDcooor^cx>CM^;LO
COLOLOCOCq^CDCD
dddddooo
dddddddoo
ddddddddoo
o825888!8S85?.
co^coLq^LqcqcDCDLqiq
ddddddddooo
co c'i m p
ooSiqcqiq^LqLqr-.r^^Lq
ddddddddddoo
ddddddddddooo
J^GO
S
n<
^s/2
$ -
s t1
oo'-jp
CD^COCMNr-COinVj
(MCJ5coco^rcoLO'-cM'<r
c^Lq^Lq^coLq^LO
iflCfilflNO^OOOoOO
focoLO^coocT-
02
- w
15s
S Hl '5
<u^S
M Tl -g
o^>,
-^^
U r
t ^L
^(L>S
>>
g (, |
m
g |
| a11
234
o) n o
cq s oq
(i-^
cq o> cq
<Mcog>
o> o> cq
<8 2>
cq cq o>
& o>
o>
i-oc)
o>O)C0
o>h-icq s C3)
o i- co
000500
evinto
o>o)d>
ffi j- j;
oqa>ci>
50 co
O)O)
T-1-1O>O>C3)
t-COt0)1^0)
OCvllO
GO O) CO
<NC0iO) O) O)
OCCM
O> O) O>
COCO
^i
T-^o
o o
L<2
a* *o
doo
oo
odo
.s
cd
o
T'
&^^2
u'o
Sj o
odd
00
odd
odd
odd
odd
.SS
^O
odd
odd
^Ph
00
odo
odd
odd
CT
C/)
S
c
^o
<D
x:
|2
IES
ri>
i S
f n^1 1-
*5f
lifi
g?
I 8
Sg^
-i=Q-.g
O.2
c
3 3
"55t5
co
^c
co to '+=
T o 2
ip^
S
*
$^o
>,?*>>
? S
g'8
Ili II I HI p! | fi
co
.E ^
collo
1 |1
^^^
'%'%'% "5|
Q-Q-Q-
gel
S5|"
go
))cl:cj'
g^
Ititeli. Jii.ttli-Iif
S
'S
tS
<
235
i
5
odd
S5
d do
888
fei
ooo
odo
SfcS
85|
ooo
ooo
oo
CO
Q.
c
(/)
.8
&
p
g <B p
g
8| 8
III
"5 g "5
i
.g < i - 8
g
o
It 111 !
g
~
S
#i
d>
|t i f H
- m m m it i i li
88
oE
ff
SU- JH
ess
5|5
^^
S1!e Hl
2 2-s 5^S
S|S
li
^8
|eo
I ill
ff
S2
||1
f ?
s I
gfi
-t
z%
fallili
S"
la
si
.Ul.iH.lliiiHlliSJSH