Transcript
8/20/2019 9 Cluster_Analysis Schaer
1/25
1
Cluster Analysis:
A practical example
8/20/2019 9 Cluster_Analysis Schaer
2/25
2
Content
Introduction: the necessity to reduce thecomplexity
Recall: what cluster analysis doesAn example : cluster analysis in consumer
research on fair trade coffee
Discussion
8/20/2019 9 Cluster_Analysis Schaer
3/25
(…)
“Where is the life, we have lost in living?
Where is the wisdom, we have lost in knowledge?
Where is the knowledge, we have lost in information?”
(…)
T. S. Elliot
!horuses from the Roc"
#1$$$ % 1&'()
Intro
8/20/2019 9 Cluster_Analysis Schaer
4/25
(…)
Where is the wisdom, we have lost in knowledge?
Where is the knowledge, we have lost in information?
(…)
“Where is the information we have lost in data?”
Intro
8/20/2019 9 Cluster_Analysis Schaer
5/25
In order to *o from data to information to
knowledge and to wisdom
we need to reduce the complexity of the data.
!omplexity can +e reduced on
, case le-el : cluster analysis
, on -aria+le le-el: factor analysis
Intro
8/20/2019 9 Cluster_Analysis Schaer
6/25
Cluster analysis can get you from this:
To this:
What cluster analysis does
a b
d
ef
c
8/20/2019 9 Cluster_Analysis Schaer
7/25
!luster analysis
*enerate *roups which are similar
homo*eneous within the *roup and as much aspossi+le hetero*eneous to other *roups
data consists usually of o+/ects or persons
se*mentation +ased on more than two -aria+les
What cluster analysis does
8/20/2019 9 Cluster_Analysis Schaer
8/25
Cluster analysis
*enerates *roups which are similar
the *roups are homo*eneous within themsel-esand as much as possi+le hetero*eneous to other
*roups
data consists usually of o+/ects or persons se*mentation is +ased on more than two
-aria+les
What cluster analysis does
8/20/2019 9 Cluster_Analysis Schaer
9/25
Examples for datasets used for clusteranalysis:
socio,economic criteria: income education profession
a*e num+er of children si0e of city of residence ....
psycho*raphic criteria: interest life style moti-ation
-alues in-ol-ement
criteria lin"ed to the +uyin* +eha-iour: price ran*e type
of media used intensity of use choice of retail outlet
fidelity +uyernon,+uyer +uyin* intensity
What cluster analysis does
8/20/2019 9 Cluster_Analysis Schaer
10/25
Proximity Measures roximity measures are used to represent the nearness of two
o+/ects
relate o+/ects with a hi*h similarity to the same cluster and o+/ects
with low similarity to different clusters
differentiation of nominal,scaled and metric,scaled -aria+les
What cluster analysis does
m
d#yiys) 3 45 6yi/,ys/6r71r
/31
y 3 -ector
is 3 different o+/ects
/ 3 the different characteristics
r 3 chan*es the wei*ht of assi*ned distancesthe calculation of the distances measures is the +asis of the
cluster analysis.
8/20/2019 9 Cluster_Analysis Schaer
11/25
Two phases:
1. 8ormin* of clusters +y the chosen data set % resultin*
in a new -aria+le that identifies cluster mem+ers
amon* the cases
2. Description of clusters +y re,crossin* with the data
What cluster analysis does
8/20/2019 9 Cluster_Analysis Schaer
12/25
Cluster Algorithm in a**lomerati-e hierarchicalclusterin* methods % se-en steps to *et clusters
1. each o+/ect is a independent cluster n
2. two clusters with the lowest distance are mer*ed to
one cluster. reduce the num+er of clusters +y 1 #n,1)
9. calculate the the distance matrix +etween the new
cluster and all remainin* clusters
. repeat step 2 and 9 #n,1) times until all o+/ects form
one remindin* cluster
What cluster analysis does
8/20/2019 9 Cluster_Analysis Schaer
13/25
Finally
1. decide upon the num+er of clusters you want to "eep
#decision often +ased on the si0e of the clusters)
2. description of the clusters +y means of the cluster,
formin* -aria+les
9. appellation of the clusters with catchy titles
What cluster analysis does
8/20/2019 9 Cluster_Analysis Schaer
14/25
What cluster analysis does
!luster (!luster !luster 9!luster 2!luster 1
8/20/2019 9 Cluster_Analysis Schaer
15/25
Practical Example
!onsumers and 8air Trade !offee #1&&;
8/20/2019 9 Cluster_Analysis Schaer
16/25
Consumers and Fair Trade Coffee
!escription of clusters:
!luster 1 #11'=): >self,oriented fair trade +uyer?
!luster 2 #19'=): >less ready to ta"e personalconstraints?
!luster 9 #1$2=): ?less en*a*ed a+out fair trade?
!luster #922=): >intensi-e +uyer?
!luster ( #1$;=): >-alue,oriented?!luster ' #('=): >does not li"e the taste of fair trade
coffee?
8/20/2019 9 Cluster_Analysis Schaer
17/25
Consumers and Fair Trade Coffee
!escription of Cluster " #""$%&': (self)oriented fair trade
buyer* :
Searches satisfaction +y doin* the *ood thin* Is not altruistic
@uys occasionally
Stic"s to his con-entional coffee +rand
i*h le-el of formal education8reBuently reli*ious #catholic or protestant)
8/20/2019 9 Cluster_Analysis Schaer
18/25
Consumers and Fair Trade Coffee
!escription of Cluster + #",$%&': (less ready to take
personal constraints*
States that >fair trade coffee is hard to find?8eels responsi+le for fare de-elopment issues
@elie-es that fair trade is efficient for de-elopin*
countries
Is less ready to *o to special fair trade outlets@uys con-entional coffee
Ci"es the taste of fair trade coffee
8/20/2019 9 Cluster_Analysis Schaer
19/25
Consumers and Fair Trade Coffee
!escription of clusters Cluster , #"-$+&': *less engaged
about fair trade* :
8eels no personal responsi+ility with re*ard tode-elopment Buestions
Doesnt see the efficiency of the consumption of fair
trade *oods
The only thin* that can ma"e him chan*e is theinfluence of friends
Is older then the a-era*e fair trade +uyer and has less
formal education
8/20/2019 9 Cluster_Analysis Schaer
20/25
Consumers and Fair Trade Coffee
!escription of clusters: Cluster . #,+$+&': (intensi/e
buyer*
as a+andoned con-entional coffee +rands
as started to +uy fair trade Buite a while a*o # 9
years)
Shops freBuently in fair trade stores #and not in or*anic
retail) Is ready to act for fair de-elopment and tal"s to friends
a+out it
Relati-ely youn* with low incomes and hi*h
educational -alues
8/20/2019 9 Cluster_Analysis Schaer
21/25
Consumers and Fair Trade Coffee
!escription of clusters: Cluster 0 #"-$1&': (/alue)
oriented*
To*ether with cluster hi*hly aware of de-elopmentissues
Ready to act and to constraint consumption ha+its
@uys for altruistic reasons
i*hly in-ol-ed in social political actionFost freBuently women hi*hest household income
amon* all clusters
Gwn security is the +asis for solidary action
8/20/2019 9 Cluster_Analysis Schaer
22/25
Consumers and Fair Trade Coffee
!escription of clusters: Cluster % #0$%&': (does not like
the taste of fair trade coffee*
Cowest purchase intensity of all clustersHot willin* to accept constraints in consumption ha+its
or hi*her prices
Fost mem+ers of these *roup are attached to a
con-entional coffee +randRelati-ely hi*h incomes a*e within the a-era*e of all
*roups lower le-el of formal education
Cess reli*ious than other *roups.
Conclusion /
8/20/2019 9 Cluster_Analysis Schaer
23/25
Conclusion /discussion
Ad/antages
no special scales of measurement necessary
hi*h persuasi-eness and *ood assi*nment to realisa+le
recommendations in practice
!isad/antages
choice of cluster,formin* -aria+les often not +ased on
theory +ut at random
determination of the ri*ht num+er of clusters often time,consumin* % often decided upon ar+itrarily
hi*h influence on the interpretation of the scientist difficult
to control #*ood documentation is needed)
Conclusion /
8/20/2019 9 Cluster_Analysis Schaer
24/25
Conclusion /discussion
8/20/2019 9 Cluster_Analysis Schaer
25/25
Russell .L. Akoff, !"rom #ata to Wisdom,! $ournal of A%%lied &'stems Anal'sis (*+*) -*.
/ilan 0elen', !/anagement &u%%ort &'stems 1owards 2ntegrated
3nowledge /anagement,! 4uman &'stems /anagement 5, no
(*+5) 6*57.
Tasha""ori A. and !h. Teddlie: !om+inin* ualitati-e and uantitai-e
Approaches. Applied Social Research Fethods Series Jolume '.Thousand Ga"s Condon Hew Delhi 1&&$.
xyxy
Sources
top related