Summary (Week 1) - Quantitative Data Analysis 2 (6012B0423Y)

QuantitativeData Analysis 2

Week 1

, 1
Topic PrincipalComponentAnalysis


PrincipalcomponentanalysisPCA

Simplifiercomplexdata
bytransformingthedataintofewerdinemion
Identifierclusters variables
of
PCAisusedwhendata is very Too
manyvariablesconfusedaboutthe ofthevariables
structure

So we prepareraw databeforefurtheranalysis toconceptualoverlap
variablesandprovidingstructures
Byreducingthe number
of
PCAisused
bytransformingvariableintocomponents
therecomponenteare uncorrelatedtoeach
other
Intherawdatacomponentsare latentwhichneam
theyarenotpresentyet
component 2


B2
iii tearing
So inthecomponent1 are all variablerelated toextraversion
Thisiscalledtheorganizing
ofvariable FactoranalyinandPCA
ForthiscoursePCAis usedin an exploratory Not the same
way

ThePCAprocedure
1Initialcheck inthedatamitableforPCA
rotatiiterretation
2MainAnalyn Componentextractionand
3followupAnalyinCreatingnewrunnaryvariables

Initialcheck datarequirementsforPCA
Quantitativevariable Intervalescaleandratioscale

Strongcorrelationamongstvariables 0,30

Largenumberofobservation whenmorethan10 thenumber variables
of
Thecomponentsarecalculatedfromco variancecorrelationbetweenthevariable



Component 3variable bevariable byvariable byvariable

, Thereare thevariables11inthis case

butthecontentoverlapeFromthiswe
wanttocreatecomponents

Cause we are analysing raw data
Include all data
Nowwewanttoanalyse
thisraw data



wewanttoseduce
thenumberofvariables



Now receive correlation
a
you
matrixWiththismatrixwe

cancheckthe
firsttwoinitialcheck

Check1 YoualsoreceivetheKMOand
Inthesamplerisesufficient
AKAisthe
couplerisemore Bartlett'stest 2
than10thevariable
10 11 111
couplerise 114 more than111 Pvalue

sufficient
check3 TheKaiserMeyerIlkintest
Check2 Aretheitemcorrelated Meaner thereis a
if
Howtobeabove
0,3 on check4 Barlett'stest diffusedpaterninite
thismeantherearemany HoMatrix Identity the
matrix correlationbetween

correlation Negativeisalsookay Hy VariablesarecorrelatedVariableor a compact
to
WewanttorejectHo patternwithitarese
whichiswhenPvalue correlation
isbelow0,05 Ifbetweenasand1 PCA