The World’s Largest Open Access Agricultural & Applied Economics Digital Library
This document is discoverable and free to researchers across the
globe due to the work of AgEcon Search.
Help ensure our sustainability.
Give to AgEcon Search
AgEcon Search
http://ageconsearch.umn.edu
aesearch@umn.edu
Papers downloaded from AgEcon Search may be used for non-commercial purposes and personal study only.
No other use, including posting to another Internet site, is permitted without permission from the copyright
owner (not AgEcon Search), or as allowed under the provisions of Fair Use, U.S. Copyright Act, Title 17 U.S.C.
Clu8terAnalysisApplie'd to the NSW
Meat Processing lntlu$try
Alison Sberidan-Nethery!'
Kate oweIll
1 Department
of Agricultural Economics
& Business Management
University ,ofNew England
2The Rural Development Centre
University ofNew England
A contributed paper to the 35th Annual Conference ortbe Australian Agricultural Economics
Society, lIto 14 February 1991, University of New England
Tbeauthorswould Uke tothatik Ray Cooksey of the.Psychology Department at UNB forbis
valuable comments,
2
Inttodu¢tion
With 1im~datvlIbcreng'o
~·yproductsefb'Aa1im
indusUy, very little research has been directed lOW$'dS understandingtheJundattlental
w()rkingsofth$ec~
Thi$as:mped~t
by·products•. TheJarger$tudythat,prpmpted thisp~$
devl(Jpmnto,:Oh~Ucig
'specifically CQncentc:d with
redsingthpoblmyJ,1-uac()~vC'SfM
sector frOnt the fU'rtl·ev(abQi;)~
WithmQrethan400abattoirs andslfiughterb(>uses J)peratif1~'\jnAsl.w
oot
possible to examine each wOrk$tothendr{lw .~pclusion
abouthe:lscr~
Foctlsing
on a sample of .tums was the only vjableproc~
With lilllited'1esearehfunds8vailable,
onlyth~wrks
in 'NSWcouldbe considered. HQwto.setect theaproi~fms
.to
study in order to capture the:important feanrres of'tbeindustry is acritica1issuewhenuslng
case..studiesas a researrb 'tool. In this case, the problem w8show tQselectabattoirs.in
NSW that capture the irnPOrtant (eatures of the totill by-ptodJ,lctsse<;:tor.
Whentllckling thisissuc" cluster .analysis seemed an appropriatetool.to use. Cluster
analysis isa generictetm encolllpassingarange oftechnique$ thatse¢k ,tQcla~sify
:8 set of
entities into a number ofctassc$.such ,that the entities withnc~sa.ve
:gteatetshnilarity
to each otherlbantheybavetoentities in.otherclasses{Everitt, 19.81). ,Applyillgcluster
:ana1ysis .to tbethe abnttoirsinNSWresultedin .soluUonswhich'initially·appeai'edro be
railyobustndvepg~
However, during the Malysi's certainsu~.o
light that were cf)t1sidered wortb.examiningintJlQre detail. These ,related·tQ the different
fonnationrules that clustering algorithms apply ingeneratlJlg .clUSfers a.ndtlle concept of
similarity. Both issues are discussed 1n ...lcjilperand possiblc·sotutioos.offered.
TheClusterim: Concept
There are essentially 5 steps involved lnany classification procedute ( Vogel, 1975).
These are:
*the selection of tbeparameters to be used for the classification
• the choice of the· similarity ~su",
* the choice of the cluster algorithm to employ .
*.tbedeelsion as to 'tbeappropriatenurriberof clusters
an evaluation ofthe usefulness of·the resulting classification (the clusters)
*
As noted in Bventt (1981,p.t03) ,;. iti$g¢neratiy lmpos$ibtc .prioritQanticipate
what comhinatlQli of varibls,ntyme~d
chtsterln:B t(:Chniques ~ .likely to
leadto.interestingand Infonnativcclassifi¢atiol)s".;()(ten 'therl,anysi$~
i nvtllves severa1stages:atwbith .ther~
mayirttcxV¢netQ makc~gt$o
variblC$,.sect'~1my
tQcusoo4ifrerent'sub-sctsOfentiti.;s.and
'make othertllinot; ~justn1er.
,tQ the 'C1us~gtQ,enr.habcl$
inu$CM
groups. It isin)portant then. ,lhatc~begnis¢ur
ofcluster .anlysi$~
~ydoing
$O.nth~user
iSfarless li(e~yt)
make tbeJllisWce oi:tCfYng~
cl~tesouin"
,(Aldendetfer ,&;alashfield, :1984,p.t 4).
TheCIJoiceand ;MeasurementQf Variables
The initial choice of variables to include in clustes:analys!s isajud~ntbyhe
researcherastQ whicatrbues,ond~
important (orthepllrp(>Sesot.classification.
AsstJch, thefinalcs~ovdr1!
in :asense,arbitraryasthe classe$ prod~
bytheclusringpQCd~a
natunPly. greatly influenced 'hythe ·Variablesemploy¢<t
Theclustrbod,f'()nyv~wipJ"
and US¢s"
(Man, 1989,p.27). 'Irtthisstudy, t1eprimayQbj~v
r..ticlust¢nn,g·wasto id~nfy
classesofabauoirsinwhicbthemelllbers ·~rela.tivy
,a1ikentbrPQduco~C$.
The infonnation available concerning dle61licensed abattPirseummtlyopenttingin
NSW was fairly limited asproductinfOl.S~
comi11efCiallysensitive. cot
theinfonnationlliatcouldbe attained relating to the works"thefli~g
vari~les
were
c()nsideredrelevant for sorting abattoirs 'into 'relatively homogeneous groups.
YARIA'BLE
Volume of throughput
Type of license
STATE
high
medium
loW
Export
Class 1
ClasS 2
Rendrig~cts
Presence
Absence
Species Slaughtered
Part
or a Company
Cattle
Sheep
Other
Yes
No
4
'I'he'Qataconcelllingthesevariablcs forthe,abAttQirsin 'NSW~
obtaitledCrom'
perS()nlllcQtnmllmcation with staff ofthe !NewSouthWales Meatlnd.ustry ,Authority
w.eU 'asinformanon
(NSWMIA.)andteJ~prmofP,byus:iEg
Plbised;nth~
'NSWMIA,'s Anlluat'Repon,(l989). 1tedrmiOl;oa¢hf~vbs
isgenenllly se1f~xplJUt()'.Wih
respect to the type ofUcensesheld"tbe ,sen~
distinction betwec:nthe three 'licenses isconcerow With ~ :healtndbygi~s
littit
:th¢ WQrksbave to m~t,
1berqu~mhts
for the :'axpon li~n$C.a1'
roorestringentthan for
the Class 1 Ucense, wblch in. tum are mQrestrlngentthan fortheCtas,s.Z liceJ~.
The'othel'
species slaughtered'arepigs and goats.
AU :thevariables w~tead·
di¢hotmus,$1:S(W~e
~'lQtOpin4caJe
the presenccQr·a,I)$enc¢' QfUle fcatllr¢, respectively. For the Yariabte$"tvQlurneof
thrOOghput" 'and "'type ofUcensetJasQUtlined abov¢,the~
were .more· ,thantwC)states
possible. However, byincl1l4ingpnly ~ostae
for ~ch
variable, then if bptll W~
coded
O,by implication the 'variable must ~·in
the third state. Fotinstante,byinc1llgmg.onJy
high ~d mediuM as:the states possible for "volume .ofthroughput".andifboth.are cod~'O,
theabattoirrnu$thavea lowtbrol1ghput Similarly, byinclU;dingonlytbe;$tates,JlxPQtt and'
·Class 1for tb~ variable "type of1icens~'.adbtu()y
illlplicab"p ,the abattOir
has a Class 2 License.
Ifa works proCesses Ii lowyolume of.thrOugpasl'CI~
rendering.(acilitics. ~laughters
tattle IUldShePtbunpors~i
to I company, then it WQu]d be coded-as follows:
1 U~nseild
SJKtd~sl\()beo
wbereH=Higb throughput
M~'C(iumthrogp
X =E'gpottUc¢n$e
Cb=Class 1 Ucense
RF:endrigtlC~
;flndFie(198,p.?)rovau~hckst
CA=Cattle$laughten:d
SH=Sheepslaughtered
OT= Othetspecies slaughten:d
Co=Part of·,acQmpany
:&ch ofthe61abattotrs werec.oded in this manner 10 create. the data set. 'Tabachnick
the researcher should folIowwben
:s
c.tat.. ~t
~in'l
'an1lY8es~
t()"~nsu¢
ther~,
Catrying:~,heck$
61abtoirs~npve()ul:$
,~y
n()proble$With'd~a
WiUQtdy~lY'af
,fQr~batid$C.hc
scxei'dng~t
'2of,~
•.. At;.~
.2ab~tQirs,lughe[ncf
,popU}atlonthC:Wget,$tu4yiscoficerQtdwiili. ,~$Ower
was roadcupof S9 abattQh-s wblr9clichototnQ\l$ VAtiJlb~$.
wcrenotP~
Tbe dam :set~aing
ot~P
del~trA
entinC$,;lS sinPlator dissitllUar is the,b!lSis fot ~lpSsifcaton
To ~gtU
Simlartyhowev~.cnbd,¢(L'pQ1fsJ$
whether entities are ,alikcQrnot atikerbuns~
and implentd"J'~b
,tjew~y!fbscop$
(Alder1t~,6G
,81e¢xl'ressed
Blashfield, 1984~
p~17).n<Uom
panicularcircgmstallCes with wbichpne isd~b1g,
tbeexptessions iorstnla~cvy
greatly. As ,clus~ringha
l>e(:ome increasglypou~,dJPt
n.:~h"
number oftecbniqushavm,dlp~Wyw:
to reflect particular cireumstan~,
A Vital.issue to'~adresWbn;.cl)1mgCupw
analysis,tben,is what.simil{Ujty mea$~bold
beu,~togahdrQflikncs$
betW~J'lh.nisw
which the r¢seachi$.~ng
div~felso
.
Wben Ute data are dicnotQlT14)u$, .sin'lilarlt)'and .dissimilaritybetwetQ'tWo cases .i$
repsn~itb
'fpresnc·tQ1"ab~
.orJ)~met·
depi¢,~jntb
folwingt~
way rss&:ietiooiable:
CASE 1
__~l~+-
CASE.2 1
a
b
__
.a+b
wh~a
o
=joint.presence ofafeatllre
b =~bsenc?f
f~ture
.for Case 1 and presence ;offeature for Case 2
c=presenceof feature Car ease 1 and absence pf feature for Qu;e. 2
d= jointabsenceof.a feature
p:;a+btctd
A numbetQr~a.ivsA1':
avilb~t()re"Psngh
cndties
on
sim1$tybc:w~n
when 'dJeatricboQf)(usWl$1987~Hwv,I
·are$ltnpy~.iQs
~bl$ic
measu~.of
association:
td.n:apl¢matdling
a+d/a+p"f1:+d
a/a+b+c
b+c:/a+btc+<l.
Ja~crd
binary . euclidean
aoth &irnple matcbin8dl~resyf$QCP
~omtnQUiybews.
~ydiferonl)
~ltQfhej9in,absp
2ulattnbute.Wilh sirnPlernatcbing, joint absences tndica~sblry.
Ja¢~'scoefint.
which 19noresth~
joint absen~
of features, 'was devl~pt()mcsI'Ciaryn
9Qe$
wbereitisinapPJ'()priate, t() con$idertw~lsamhJ.b¢u
:1x>mlttck MtUtrlbute.
Binary eltUd~is
adistnce~mf.("roU,ly
tIultit
focuses on .thenumOOrof Jltrib~haw()
.entities (lQnot :haveinC>fJ1l'(~Wtr
the
res~h
'wisbes to focus on sbl11'ed ornon-sltaredattriPllfesd*nds(){l tb~ natUtePf.lhe
problen~ig
aQdressed.
Fortlwpm:PO$es()fhi.a~
thesimpMng·~cwado:q
all variables sil'llilarity WflS lrnpUclt inJointapsences.
In chQi~ofalgrJm,te
selecting representa!ivetirms sJ"etdwohigHly~rabcu
sbould.peF~t,ahmi$nry
distance.(lr:y)bwgQup$~Fom
proposiijc:m thatclustenpg·PtQvidesan aYentJcfQf
tbatana!gpritbm
between entiticswidtin gro,upsandUte
thabden~y
thelarg~myof
.to find :tighbypel'$rc~us
we~
~S6rUbmsavAletQ
CPllsidercdthe Dl()St
.$;ls$imllar.aspPS$ibleoyer
appropriate. ThenatlrCQf:~
clgsterjmpih.an~
all dimensions (variables). 1bis.tendencyhas beencriticiSC!d 'inSQn)e .discp1~$:f«t
persi$tanc.3Plygh~b
n!gfU'<Uess ()f meutldel'lyjngshape' (Everit,p~957)
For tXQJll»le,hypeJ."Spheres .may be considered inappropriate where entities can have a WicJe
:rnnge pVcr one.()t .more dimensions but $~UbeCQnsid¢ml-.That
b,real.cl\l$terstnay
be~l()fptdnhr
tha,nypmericJ.WbU:$~
is ~nsiprote
this$wdy iim4 indeed conceivably lnmany econonUc applications, itroaY' ~ con$id~re
inappropriate in some instances.
7
'The $~Qnd,
·~m<)re.in1ly,atbu$
~lu,tg!"Qjhrs
sbou1dgelrJtCpima$c>~
'WbU~a:JulmcrQf
(Bverit~).wcly·b
~$Qritbave@'dy¢lop
applie4rodi.chotOJll()US '~
.di~ve
.~
,~.Jijng:techmqls'(avb¢
;TWQ.,e~«:pnloUtc
Wi$luu:t(1987)suito:of.pro,nuns)"
..,orit.hms~·plbu
'In th~inlW.Sagofmy$sbeQ
w.ith (idler$uQ;po~1wm
$Olllti()ns we~At()ds
dn>pe.c:ltiQrh~·;a1ys
:nemotb~icdvhQu(W$Jr,
p~790)
anlUy,is whe,rHb~fint
data.
:~ntedm
As ·... ~tClT3iveo
whict~
~IAluinth
dle·.ijnal ~tiQn
•.9f,thepa~wuUC4
.JO . ~,
optimngalQrdu1,S~"9J.
~PQS
~
~lp$B;staeY
~lQtions('.MPI$gJ1rhTeu"pA
M4sqbject tbe~
:JQCAl.Qpgmum only~
'T()incretl$ethe lU<el!boPdtluttaglpbilloppqlumhas'*n ~h«I'requStcpmJon
Jrom a nurbe()fs~gpomt
($¢lec~itb.nud)Qyfx>mp
thrOugh·otb¢r
'~gQrithl$).
lt1iaS9uonJsB·grhm.~
gent1il:p~
,,1.,$' $QftW~1ilaons
p~ltJde4
$Cttle4 ,pponfor thi~$udy
theJlsof,mi$ra~gy
~tb.apfUl
W.MtQlJa1y$Cde~Sing
cltJseriga()m~Toh
,ft nPmbet ot.$uiUb~hm'cl1
seJienlted $imU~SQlt(
d~lh"'tbaSQim
lev~rcqu,th
:re~ulting
W{IS
.,~dl
ctllstcrs<»u1d, be~gardsJ'Qutloh
,dJ~1gbtno
CQtlsPu~anpir.
'lbe~ngtU$cQsi()r
,b,'Uu;
.hasO\lUinedWocb~rupt
autbon;to ~higlydbe
inv~sjgatc:)Q
'Th~
wJUQhn~adresig
'inclustqinSa!gQriwrns anr;l·.~theS1bj
iS$uc$a5sociated with mesatriblp~noJ)'
ifc11)stering is to b~ 'fruitDlJ~noUca.\Iy$C
Witbnh~:eracCl;$1.ymodfv
.~
'tecbniqpcs .~in
pfongomJJ
lew·oftho$e
geneflll u~.
(opr
ofthese Wf,i'C consi4er.d,Th~a:
CctmRJeu; )jnk,a" method, whi«:bcluste1'$by ·tbcnJlcthat amy entityto.beincluded
intoM e~i.sdnglJWh\S
tQbe;··wlthin '.a cenainlevel ()f:~imb.uty
tQ ~ lllernber$pfUmt
,cJuster'(A:fn4~adBlhi
,p~4()thai
varince.Thsmt()Cb~
clustef$>mad~.P
of v~ry
shown.tohave atendcy()firomp,b~h¢U
similar (:as¢s.
itUes()mn~
tbeintrtlrcla$s
W0;4'§metb<>4 >is desi~tomlh"
th~
vari~wthn
Clu$t¢rnB,J).~fio(aISsbp
.whQS~
ofcluster$ i$fl$$Cs$t;d im,~tbe
increase inth¢ v~an
~combine<fvrt.p9l)
wbi¢h .~1'Qqgbly·eat
iJl~Z)Wd.
unjoJlI~$·tQhes
W~sl1¢th04_gne"Ay
clu$teJl.At each sblg.,'jn
.gevlQP$·cust~
twc> .c~$'et
hyPetsp(:rica1.nb~
~aV¢u:e
Jinqt~meb9d
Openltes1>),deriVlrlg'¢r) average, value foritbe sitPilArity
waUcas¢$ ,in·tb¢ ¢xistinS clU$ter,Ai~
Qf.~ca$ebingo"sdrltv
~joil)$
ifa.level ofsitnlladty :i$lllWn¢dU$inBtblS :avergJu~
clUst~r
panicPl~yuserJ
i~
This !~S()rithm$
insq>fU'atjngoUtIieJ'$.
The penqoi 4m;w94 depictshlurn~
fonne4tl1ey .~tePIaCdbylcO>rinsQh
centrQid.
.are. then fused~
1li.s~thOd$npouarbecQ
simple (l.4mce &, WjlU~,
distance~w.:hro
space~nrvig
aIld·JS~custer$
.an4is ~lativ¢y
iNoollctechmqUe is considen:<i$up¢rior tQ t~
However, only twO, WArd's and completnk~g.
'.~
Th~grQupswit
lbesn:u~.t
it 'ls
1967).
othet$ in ltU~in;umsJace$.
llkelyto sati$fyb,~
~uiremnt
of
tj.gbypersh~
Widlbienu:thical.clusteringtwbniqllcs, -the res.~h,mutdci
clustering.}$ mQ$tsignifiCAAtThaJ ls, 'wh~t
n~mberofclusJS;$t¢apQ1·i
gro1,Jpin~sf
tbe~a()ir$.
natunU
VariOtlS stop.iJlrueh~vbn
mi(l fast .anc1few .llI'eappUcable ~saU
.clusterssh()uld ~.
wbatlevel p(
noe~
:h~
clustering ·a]g()ritbms. BQwever,atdlc'n)inhnutn
c(mteJ(tofthed1\ta.
ilt~JCm1>ewhn
Apracul~thodfing:wbe
toJernliacus~h'$d,pm
isplottingtbeclassifi9aticmcriterionagainstllie number pfclusfet$ .tu)d.· ..asharp
nurober of.claset~(Owr,
1975)~aQwe"r,'
,this method
ac~ptne,
.stepinhJdca~
is reco$ni~u
being subjecriYan1h~fod
studies ltmust:\)epse<i cautl()s'~Wh!1eringqm
twod~sivep
to '~inplcabe
we~aplidtoh
w~reapClfot
Thefinalst~p:
·.three· ,andtbe four~lpstin1'd
~onebtwh
the Qtber belWe¢ntl}escvenandeight c1usl~r
discunglthepmbaSQC~
ina numberQf
sol"tpJ$(~eni
.Apendi,(1)~
flch$~eringpQCdut.Vaobs1Ev¢,m
witbjpdgingtbevalidityQf clustmSp~$haQi1.e
m~$()fYalidt.1ghe
clpstCll formedisd:tat of:runningavanety .of~lustc;riagJh1$
ba#d·on :diffetentMsumpti()fls, artd"onlycU$ef()m.bhji~
datA
rned19s:b~ptf
"averit,p.7S)~1b$JPQc:
algoritunsv~
wM~¢Il,andtheusg
'lincage,dtrow~
WtU'd's JIld'~mp¢
tQ yJWdate'~
liOkagcsolud9ns.
The reS\lIting ~lJstC"$difre4onwha;\,#!:gQu'
l.owlevcls,()f,fusiQll
buteonv~,g
Atbi shcrleve1.$. ThefQPr·clJst~uion
for tbeWQnf'~
~I}t«)id
a1gcdtlunswe~
identical. ~dtbc:r
\W~
()Jlly .dfer~,btwQCnJlhUs·o
,tIu'ee
abattoiQ by .Jbe~
,lllgcritlun$'andtlleaVenls¢.JinQgc8Jl4complete linkagc$Ql\luQn$(tne
three ~.irswbatdleh1fnUy"m¢c
,oftheJa~i1.w
Alg~).
As
th~rels$ucbamdifn"w.o1Q,
,tbe ~lustef$nabyWm
clusterif1g,p'QC$\V~jdo
be ,ftUpyrobUst.
'Thefour.clostell:Siven 'by Wani's ,~
pre lis~nApeQdx
MtlIysisQfabattoirswillolllybe ~lev8Jtfor
tbi$stydyif tbecl"$J~,Qi{nmo
~he principalcQaracteristics.ThedistblgWsbipg c~trh;q9fe
,~AJ;l$ter
of
~lust$·,are4i
:~h)w.
The (lI'St .gro\lP is mll4epp()f 14 }abl1tQirswhC~v
aXpOn.licellSC$.Theworlcs:in
thi$cafe~u.
~$lausb
catl~(wib
:~td$lauberig.
on~ieJ)ad
dley
Jill have ~ndrigftes.
nere,a,e afew~
wodcs:.~tJla\ghernIy
...
fcmturenot fO\lJldirtanywotlc$ just se.rvicin ~ :tl\e localma.rkct.
The .ciisUngliisbingfeatme of the second group,iJU\llt cacbof .thO· five wQdc$only
s1.au~hter$p
'1breeof the five ;wQrk$hllve·an export1i~nldbIQgJ>a:ms.
FOUfof lhe w9tkshandle alow yolunte ofthrC)ughpllt .. dleexpol1 WQ~
h~dlirg.a
volume ,prtbPJlShuasony'«e~
complet¢daniSfr9C$O~,bQ
t,he .()ther wQJ'kii.
n.e~
·.~18(lbatoirsgneh4
clu$~,
ISo! wbicb,bavea ~la$
t 1icn~
"
lbethreewOricSwith a Class .~ Ucen~
.prelncludetl ind.tsca~g()1y:U$elYhr
·a·
medium volume Qftbrougbput aJl;l\lyPlcalfeatJ,irefor Class 2. wotics.()nlytlu'ee gf the
worb,hl.tbt·S cat~orybelJg'Q
.CQmp~yandt1l;
barp!le.slJlughtet,aU species.
.'!'
,wt! largest and is .rna4e,. lPoft~reuUnig2
The final c.~oryis
~lWQrk$haO
only ~srntv()lue
Clas$2 woncs~
.Qfthro\lghpytand does .noflx:1Qngto a cprnpany,
The Vil$troajQriJyQf thc~laugberp¢i
.thQ~genly
'"
.,
Wbil~
Thj$8J'OUP~sent:,il
;,$Crvi~blgUtejJ.
.
·~four
.b¢nceth¢lQw VQl~9f.t:Im>ugbp
·onal.~et!>
clustersolpti<>n$4Usn«l1»th;tlle tpWytical.1IDd n:~
that n~4
of lhe 'wiqersttldy, t1US:$QhlriOWn()op~
~ilJnt"cge'$axsmhd.
l~
in UteclJs~hrm-)'
iJ)$Ulbn~ofsutQ"_w:d
investg~o,
Inthex~o
lh~
The in$tabllity Qf clust~
.. bignU~
.tbeprvious~n
WQrQ
clust~agorihm;d
undeyigca$tof~
tc;qglre~ns
nOtfCprc~l,he
probable
J\$S\lclt .ilQoptimal$Olutlonwoulc,Jc,qst at • :l()w~r
:lovel$ Wtl118nted
ln$tabiUty .~gi$CQ
mebrsipat.lQw~odc\SO
meniloneg .in
anUrtl~QfpobeSPsciAwhu.,
AJl4. appUcatiQnof:sirrUhnity
clu~tering.sAadwb1o
to dichotomous da~
Two SOllJ'Ce$ of instability were jdenti.f,ThmQ~
SOQrCe lies in tbe QiffCl'entJQinins rulesdtrit 9b~«=rQ!jU1tn$plyi
Unl~$
~m
,cQIltaindisUnctgrQups()f ~nti?mO¢aopf.rgc
can vary 8C¢ordingtqme ~8oriUlt'emp(»d.F)
elongat~scUr
9£ enti:~s$g1¢
lingge «)n~(f
,nearestne,iahbQurtulewUl PJDbal)'prodg~$in,e
hl1nd. ~k$to
,minbsetra~lu'vc
,ntQrctight clusters. While this problem h;t$I)e(.1invga4bymr~,
,CQtn1llQn ··J!n4·disruptlve
fQilng~ps.
J~
~bienuchal
'loscJu$~¢r
and would m~r.tendOpoPC
lnJ~eQits
siven .•u~latv¢JywPrQm
famlly),'wbi<:h jQin$.pn
W1P'ds,on '~Ol
:t\Voor
applicable ruleS·AS to wne#.pAJticuar1sQhmoldb'v~
gcm~'Jty
devlo~
The~ond$OUfc·()jstlbiyWSu
to "~ties
lnmC$im.ilfbity<:oefficients.
Tie$occurwh¢n tWo enti~s
:haveiQenucal. $hniJaritytoa ~thirdenpy
but(!iffet,fromeacb
()m~.
ToiU\lstrlltt;,. tt"te simiJarif¥l>etween the fQltowinS·~.
enudes' w~s ~culaU!dsin
simplematcbing.
Variables
~!
,CaseS
.r... : .....
'~ i.•·.· .•~.· .•· :.•. ~ .•.
i.·. ·i.·•.·. :. .•.•
10 .100 1 DO 1
11
1
,77.$
;2
3
1 ana. 3 ,;U-¢ ~t
.8$. $q)tu1leC"s~,y
.it .1$ lmp9SsUjl¢ t()
distn~lh
with which .cluslCr entity2sn9pldjoill, Thispr.edicamenf is .n<>t .ur4que to
$implematching. 'Sinc¢ mt binary f1S$QCi~tnm
·~.mputedfrQJb,
·tw~W,a)'
3sSQCiatiQrltab)e"t\C$ .(:~, Q¢Clr.eg~d,sfWhic#4Pt1'\
ti~.are
v~
.rewme
problem may not be sev~bptwhnd
ma.trixh$nubeQf~
cluster.$Od,on~y
be<:<>me bishly.\ntae,J~
next$eCUQn.tm ~tenau":Uod
for Q~culatin,bry
Ifenti~s
"
'
'
,
\
shnUarityC)efc¢jpo~'Wb!.l:
-
,
,
ov~spe$
InUle preceding$CCtionequivalent difet1~,
cases w~oneftbcdiQl1s
:idertfc.l~,ausng
otie~
in, the .$lrnUait)'~w
ins~bUty
incblster m.e~hip
'The
likelihood ofthisCQnditionoccuring isbiglleJ' .for·dichotoroous MUitban fQI c(mtinuQ\lsot
mUlti-state data .as dicho{QffiOus da~.re
limitedw two state va1u~.
Con~Qtly,
iustabilityin loweronfer hienm::mca.lclustcrsPhltions c~be,d
toOCC\li' nt<>re
frequently when similarity i$calut~
.fm,mdichotomousdata. lntb$~"o
anew
npproJIChtoclllculatingSimilcu;tyfrombinary d~whicb$
designtQ.ruc~b
li.kelihqQd of ties, is .,:ported.
(1982) and others haveexplQrrA
lb~
aplic~ton
iofalgebraic ~opl()gy.
in the
.
anlysi~ofthe'Ucur
ofsccialreladons within a groupo(individualS.ThedilUl tllatare
the .spbject ofanalysi$ indlis contt?te ar¢Qrtert .<JiCho~!1lO$Fr
.exlIlllp1e. the data ~maybe
inthefoM·of.an inC,jdec~ma\'}Qg
e~h
ind;ru~s
plU'ticipation ,in .·l~ries
Qf
·Jlcpvi.tes~Thaon
of a1g~bri¢
topcllogy Involves.analy$is of.dle~pQnc
m~tri"
wbichis formed by the ptoduc~f
the in(;idencernatrix .anditS trans~e(Pobl,
~J;m
p.219;~)ThecQ-Consd¢matrix&
indlvu1sjomtyprcJ>a~
'simplicial ~ple?t'{U1dC9rs
-
~.
.
the .frequency wbh wbichany two
urtacivy~
A toW (or column) ofthis matrix is termed a
the fre.q\lency With Which an individualjointlY,pardCipateP
12
inactivities Witb ~;tch
()ftb~·
individuals.
the ·groupinvQlvc$ 'analysis ofthe ipetwod~
of~i,mpUcalCQeX$
·Th~:identca1o()r.l
$'p~tnl:o£
!&c:>manotbr~l'Speiy",hfkJCA·
.
.interp~dasmqQfl
el~nt()f.,
$itnlmy~wo
'simplicial comple,~n:ts
the frequency with wbi¢bdlcy jomtlypanic~he,vJs
'l~\viduas.In9WPr$tch
'tbesinl~y.)
twQimlividl.UlJ.sln .~tm$·()f
Another.PQssibility t which does ,nmappear.tc) baYel>e¢n cxplOJ'ed. J$ .t()rea,h~
s.implcaoe~;nt
am~ure·of
similarity. .In otherw<mJs,iW; panem of.~
c~dnew
eacb :jndivu~
and .llll othc:i,ndvjua1$~mesb
of
similarity..lf .two inQividllalS possess identical Simplicial cQfnple~sth
similarity
between dto~
two individuals is exact, Therno~
·the :simplicialcpmplexesdiffet:, ·lhe.P1Qfe
dissimilar are two individuals.
The employn~f
the simplicial cornplex;asa Jlletl$\lfeofsimih\rity wbel.~ta
are
dichQtol1l()1,ls has the ~fecl
of diminishing the :likelihpod of .tie$inshnilllrity~ Th~ reduction
int~arse
from two source~Fit,
the value.$of ~simp1cale"Mut$
rather than dichotomous. Second, use ofthec()mpl~x
involves,implicitly,comparisolls
between entities .acrossattributes'anda(:tOss all entities in the data. CQnvepoa1<~urs
of sirnilarityinvolve. comparisionsbetween entities acrostibu~pnly,
simlartyh~
To operationalise. the uSCQf the simplicial complex as a·basiformcun~
followingsirnilarity measure was developed:
i,j,k = l, ••••• ,n
Where» is the number of individulllS Of cases, and thf'l :"IJ denote¢lernent Jor
simplicial complex i,nor.maUsed by the primary simplex for case 1, Algebraicallyt
"ij =wij I wii
Where WiJ is element Jofsimplicial complex i.
13
Thelt$r.mg~uJ'C
is.~
dissirniJarityCoefficient.pOs&e$sil1g,b()th ~p;and
bounds,the.c()mi~
.canbc int¢rpfe~
.asnrasuring.the ¢ucUd~rtis
simplicialcomplexe$ (Everitt, p.l7),
When the coefficiealtwas etnployedto .meas~;i#ltrybwn()'$
lower
,betwetmtwo
the
nutnberof deswaignfcJltIY~U1h$mpQ
matCblng¢Qfieh~r
·'were
37 IS ties present ill the 'Illatrhc_ In ton)paris.he'w~
43desp~ntih.marx
when the proposedcoefftcicntwa&'adopted ....TedQCtion,8pr~.
ntc ¢qefficient
clearlY' offers 'pronuseasa means ofreduclngthe effect·()fties.OIl clustttinstability.
To ~stwbehrcpoPQdmau
'irnprovesstabiUty .~tw¢en:a1gorihmsc
simihlrity mamceswere used to fonclist~'ug
W~ntsiloda
.completinkg~
Instability, measured intenns OfCOOll1lOllClassification ·()fCtlSeS, was'improvedtU~
siX
cluster solution but reduced at .the fouticblser;SQn~
Atthesix cluster sotution.73 .~
cent of cases were ,placed in thesameclassificanQn.using the proposee coefficient while 69
per cent ofcase$ wercplaccd in "he sante classlncatiQn when botbalgorithms w~ap1ied
tQtbesimpletnatcbin,g coefficient. At the (Pur clustersoluuonthe results Were '61 :and78
percent for theprosdanDlcig~f:uvy.
The decrease instabilityofclllster ,SOllltionsat' ·broaderclassmcation levelsusirtgthe
proposed coefficient probably arises frotntWo $Ouree5.. ·First,thc; lower the level of :the
hieran:hica1 solution,thernoreimportant istheroleplay¢4 ~y nessince differences in
joining rules (that is, clllstering atgonthms) are ,tllQte likely to generate dj.f~nC$,i
cluster
membership. Second,the range ofstatcvalues pfconverttionalbinarysimilarity coefficients
is limited to the number of variables. In conU'ast .the range rorthe proposed.coefficient isa .
function of: the number orvaibles~
the number oreases a14thediv~y
of simplicial
complexes. Therefore. when a conventional coeffickentisUsed different clustering
·algoritbmsarelikely·toconvergemore quicklytohesamrng.~
levels
in the hierarchy• This behaviour emphasises .the importanceofchoosint; the C()rrect level!'n
the hierarchy as the appropriate c)u$terlngsolution. Ittil$OhighUght$ .the need for more
robust optimisil1g clus~ng
algorithms. to be developed toassistreseatcbers in .tnJOOngthis
choice. .By way ofexamplet the optimising procedure ": 'monothenc divisive clus~ng
(mentiod~1r)
was ctllployed()Ilthe data. The treatment ot the data by UUs,procedure
mostclosclyresembles. ofall the Clustering ·procedures. 'the 'manner in wbicb datr~
treatedbyilieproposedcoefficient.Thisproceduregenetated anoptin'lum .solution of.seven
clusters 8i1d 90 per centofcases. were placed in the same classifications that are obtaine4in a
seven cluster solution usingWani's·metbod.For .complete linlcage. a70percent agreement
14
with th~$Cve.clusr()ijo1d,
whiJe',tconCOIabW~.8d
,cQmplet$Oudons.a~Wrw
76petcm.Thi$o~Wa:rn
agreement betw~n
Want's and.complete WheJl '·$Cvcn.clustet$OllltiOtt'ls .g~neratdJsi
tbe sinlple.tn;ltching coefficient
Th~sirnlatyuepod
$hows.pruUea:~(
:filt.MUringsimUaritywhcm na~
,art)dicho(lU~Fune
work iStJ~yofneh
conceptS pre$Cnteqhere .andto cxpalldtheiJPpliclitl(:m .Qf,Jhetoeffici¢nt. AdditionalwQ[k .is
being undenaken todevelqpan optimising clusterinJ:a1gQbmfo;~
witb~.
~ficten.
Thepurposc ,of~pIYi"g(luster'anyhNSWm.dw
·to establish some· meansotcategorising ·abattoirs into relatively 'hQnlg~JOqsroupF
~ach
ofthes¢gtoupsasarnplefinn could'ilien.bednlwnwbich would capturc thtmng¢.of
workscUITentlyoperadng.Jnso fat .~th¢
cl\lste1:'S,genetatedinthemidalanalysis ~
tairlymbustand. inlerptab~
this objective wassatlsfied.
:Whiletbe four·clu$tet' $Olption satisfied . bo~the;ma1yicnd·,sOur'q\
of.theWidc:r .study, ·.UUSSOlutionwas not optitnalmtbafitdiCl ootrepre$¢nt'lhe probable
tnaturaltclusterstllatexist in the .data. As ,sllchanoptimPl soJutionwoUld existatalbwer
level in .~clu$tering
.hienuchy,the instabilityofsOlutl.c)nsat Jowetfuslonlevels. warranted
,iny~stglo.Twurce
'of .instability were identified.thcdlfferences injoining.ndesthat
clusteratgoritbms apply in 'fonnil'lgclusters,andtbe presence ·pftie$ .i!'l'tbe data. Thefonner
source has been thefoc,s:mur~abli1
ofthe issuesmvQlved ,jsunUlcely
for$Ome time. It is, therefore, largely left tothereseareher toheuristicalJ'Y justify their
cooice()f.atgorithm. Withrespcomaungbywd~()l,
'a,
'new:approach l()cauting$r~
was presented. The attributes Qf theprPPosed
simiJaritycoefficient largely Qve~m
the problem 'wjthties.Use oftllecoeffieieotis the
subjectofnil,graChp~
,it may be of value in;lddressinganumber Qt
other iSsues con~rjgim1aty
De$pit~hQsorcml1gS
;greatpondliqu~v
t~hniqle
conep,tsflq$~y
discuSS¢(J intlttspapetthe<authors feel cll,lsfet .analysis has
economic lUla1ysis. .Apparent,lhopgh.is the ~toaplyhe
witbcaudonand an urgent need ,for furtberinvestigatiQn.intotheproblemsand
apply :in economic analyses.
.IS
WAR]) '.8 CO:E;FFICIEN"r
CENTROIDCOEF.FICIENT
~o.
O~1·-r
.......-.~tIi'
-+-'"!""P_-_ _____--,.-_ -r---,. .-....-- ----r-- .----r- -r--..- .,....-I
.0
12
Cluste-rs
16
CLUSTER!
1.Blaynf!yJ\battOir P/L
2.Cudgegollg (Abattoir) County Council
'3~GunedahSirCocl
4.Metro Meat (Moss Vale)
S:R.J,(]ilbertsQn PIL
~WiflgharnAbtoPIL
7.T11eMidCoastMeat eoI'lL
8:NorthWestExpotts ,P/L
9. NorthemC()o!opMeat CQ Ltd
.10. The' Aberdeen Jleef Ct)
11.(Jgyra M~tPacldng
I'lL
12.Riverstone Metit ·CQPIL
P/L
14.MetroMeat(Wagga)Ltd
13.Lacht~y,Me$
CLUSTER 2
1S.AustralianSuperiorMeatPackersPIL
16.Australian Specialsd~
MeatProductsPIL
17~Fletchr
IntemationalBxportsP/L
27~StadmnPlL
S9.SlPayMeatCo.
CLUSTER 3
18 Beers Butcheries :P/L
19.A.l. Bush &SonsP/L
20~Castlergh
Regional Abauoir
21.F.ONicbols,(Abattoir}
Z2.Tamworth .City Abattoir
23. T.& J.Wadland :P/L.
24.CowraAbattoir Ltd
2S.PJP'rish Mcmt Supplies
26.G.M.ScQUPIL
28.S.Motrow &. Sons
29.Griffith Abattoir
30~BurtaJ18ong
Abattoirs
..~
17
'3.l . R.NTaylorP~
'32~\1ndra
Abattoir
PIt.,
,~3JI!l$tingMeaS\py
34~I.RB\lrnt
3S.EJenMelts P/L
39.WoUandillyO:nttJ!l. :Killing,O>-Qp,
CLUSTER 4
36.()llJUblgaiKillingCentre
37.WandaeIPas,tora} Company
38.~yPorah
& Sons
,40.Swans Butchery
41~).&
S~Af1ick
42~DeringuaMts
43.Wyalong Meatwotics
'4.F~ericktOn
Meat Servi<:e
4S~YoP
46.BablaraaJ'/L
47,.MurwU1im~J\bat()P/L
48.Moore l/iLatimore
49.CunnynghliJl1Btos
SO.Tancred &. Conslllble
Sl.SPulbernJUverina 'M~tSuply
S2.R.G.Uat1oW
S3!C.p.Ba}1 Butchery & Co
54.C.0. Perry&Cp
S~R.&N
Stenhouse
S6.'P.R.& R.N~Andrews
,57.MorgtUlsllulchery
58.Macleay Valley MeatS
.P/L
18
~Retrnf!$
Alden~r.MS,&Bashfi
RJ<:.(1984), quslerAnalxsis.$agePiJ1)licatlQns.
.
.
J1evetlyHills.
Clifford,a:r. &. 'Stephenson,W•.(1915), ,An IntrqduQtipl1 TQN\lmerical Classification,
Academi.c l:'re~s,NwYok.
..
.'
. ,.
'Donen, p~
(1982)~
'Orlthe J)elineationof 'Sm;tU GroupS~ct'
,in Hud$On;H.&
J)aJ8. JQ$ey..;SClS$inc, ·San. PnmCisco
Associate.s. CJasifyl~oR,
Everitt, ,B. (1981),Clu$ter AJla1)!sj,2nde.tPrNwYok~
Oower,l~C.
(1911),A Gen~CoficJ
Gren.PE~,
J3iQmFtrC$.')7,8S~2
ofSimilarity and SOm¢ofjtsPr~e,
TuU, D.$~&Albau,G(98)
Prenti~.;HalmoEc:
Res_FQrMykc;tin~
New York.. .
.
.,. "
..
pWi§ions.
...,
1arcUnc,N.&SibSQn, R. (1968), The ConstructiQn QfaiertU'Chcnlo~Hb¢
Classifications,C9'UpUl$!rloumaJ. 11, l17...184
Rum~l,Dl
'R,.J. (t970), 'AppJiesJFacto[
AnaJysi~.Notl\Ve¢,rUvCf$P;Q
... ' ., ..................
SJl~tbP,H.A&
Sokal R.R •.(1973),HumedcalTxQnoDy"W:~Ftr .
Fra~i
Tabchni:k,B~G.&FldetLS(98)J1sMvjyrgC2
QfSl J\ndJews.Edinburgh
.
4 CO,SJ\1l
...............................' .....
RpwPDblishers,New York.
Wishm,]). (l987).ChJ~tanUr
'. . .
' . ' ...
.
.
.
«lb.·a(U'p:t~
Mannal. 4th Edition, ComputingybQratc>ry, Univ~qst)'·
,