class factor analysis
TRANSCRIPT
-
8/14/2019 Class Factor Analysis
1/21
19-1
Presentation on Factor Analysis
-
8/14/2019 Class Factor Analysis
2/21
1) Factor Analysis
2) Use and application
3) Statistics Associated with Factor Analysis
4) Conducting Factor Analysis
5)Applications of Common Factor Analysis
6) Example: Life Insurance
Content
-
8/14/2019 Class Factor Analysis
3/21
19-3
Factor Analysis
Factor analysis:For data reduction and summarization.
Factor analysis:Interdependence technique: No distinction between dependent andindependent variables.
Factor analysis is used in the following circumstances:
To identify underlying dimensions or factors.
To identify a new, smaller, set of uncorrelated variables to replace. (Reg. &DA)
For MA:a smaller set of salient variables from a larger set.
-
8/14/2019 Class Factor Analysis
4/21
19-4
Application in Market Research
Market segmentation : Economy , Performance,Comfort.
Product design:Brand attributes.
Pricing studies: Characteristics of price sensitivecustomer.
i i i d i h
-
8/14/2019 Class Factor Analysis
5/21
19-5Statistics Associated with FactorAnalysis
Bartlett's test of sphericity: Identity matirx Correlation matrix.
Communality. Eigenvalue. Factor loadings.. Factor loading plot. A factor loading plot is a plot
of the original variables using the factor loadings ascoordinates.
Factor matrix. Loadings of all the variables on all thefactors extracted.
Factor scores:Composite scores-- on derived factors.
S i i A i d i h F
-
8/14/2019 Class Factor Analysis
6/21
19-6
Kaiser-Meyer-Olkin (KMO) measure ofsampling adequacy. An index used toexamine the appropriateness of factoranalysis.
Percentage of variance. Thepercentage of the total variance attributedto each factor.
Residuals .
Scree plot. Eigenvalues Vs. number offactors in order of extraction.
Statistics Associated with FactorAnalysis
-
8/14/2019 Class Factor Analysis
7/21
19-7
Conducting Factor Analysis
Construction of the Correlation Matrix
After 2 tests-Method of Factor Analysis
Determination of Number of Factors
Determination of Model Fit
Problem formulation
Calculation ofFactor Scores
Interpretation of Factors
Rotation of Factors
Selection ofSurrogate Variables
Varia.&SS
-
8/14/2019 Class Factor Analysis
8/21
19-8
In principal components analysis,
(1)The total variance in the data is considered.
(2) PCA is recommended .
when the primary concern is to determine the minimumnumber of factors that will account for maximum variancein the data for use in subsequent multivariate analysis.
In common factor analysis:
(1)The factors are estimated based only on thecommon variance.
(2) Communalities are inserted in the diagonal of the
correlation matrix.(3) This method is appropriate when the primary
concern is to identify the underlying dimensions and thecommon variance is of interest.
Conducting Factor AnalysisDetermine the Method of Factor Analysis
-
8/14/2019 Class Factor Analysis
9/21
19-9Factors
Smaller no of factors : To Summarize theinformation.
How many
1. A priori: researcher knowledge.
2. Determination based on eigenvalue:- More than1.
3. Determination based on % of variance:at least60%
4. Determination based on Split-HalfReliability:Only factors with highcorrespondence of factor loading across the twosub sample are retained.
-
8/14/2019 Class Factor Analysis
10/21
19-10Factors
5. Determination based on scree plot :.
1 532 4 60
3
1.5
Component number
E
I
G
E
N
V
A
L
U
E
Factors Eigenvalue
1 2.731
2 2.218
3 .442
4 .341
5 .183
6 .085
R lt f P i i l C t
-
8/14/2019 Class Factor Analysis
11/21
19-11Results of Principal ComponentsAnalysis
Communalities
Variables Ini
V1 1.V2 1.
V3 1.
Initial Eigen value
Amount of variance a variable shares.
Total variance associated with the factor
R lt f P i i l C t
-
8/14/2019 Class Factor Analysis
12/21
19-12Results of Principal ComponentsAnalysis
Extraction Sums o
Factor Eigen valu
1 2.732 2.21
Factor Matrix
Variables F
V1
Contains F Ls of all the variables on all the factors
extracted
19 13
-
8/14/2019 Class Factor Analysis
13/21
19-13
Rotate factor
Rotation: Factor matrix is transformed into simple one thatis easier to interpret
Rotation does not effect communalities and % of totalvariance.
Types of rotation:
Orthogonal:
Varimax- minimize no of variables with high loading ona factor
Oblique:
Rotation changes % of variance accounted by eachfactor.
Rotation Sums of S
-
8/14/2019 Class Factor Analysis
14/21
Conducting Factor AnalysisInterpret Factors
(1) A factor can then be interpreted interms of the variables that load high
on it.
(2)Another useful aid in interpretationis to plot the variables
19 15R lt f P i i l C t
-
8/14/2019 Class Factor Analysis
15/21
19-15Results of Principal ComponentsAnalysis
Rotated Facto
Variables F
V1V2
V3
Factor Score
Variables
V1-Prevents cavityV2-Shiny teeth
V3-Strengthen gums
V4-Freshens breath
V5-Prevention of tooth
decay is not animportant thing.
V6-Attractive teeth
19 16
-
8/14/2019 Class Factor Analysis
16/21
19-16
(1)A factor can then be interpreted in terms ofthe variables that load high on it.
(2)Another useful aid in interpretation is to plot
the variables
Conducting Factor AnalysisInterpret Factors
19 17
-
8/14/2019 Class Factor Analysis
17/21
19-17
Plot
1.0
0.5
0.0
-0.5
-1.0
Compo
nent
2
Component 1
Component Plot in RotatedSpace
1.0 0.5 0.0 -0.5 -1.0
V1
V3
V6
V2
V5
V4
19 18
C d i l i
-
8/14/2019 Class Factor Analysis
18/21
19-18
The factor scores for the i th factor may beestimated
as follows:
F
i= W
i1X
1+ W
i2X
2+ W
i3X
3+ . . . + W
ikX
k
Fi = estimate of ith factor
Wi = factor score coefficient
Xi = ith standard variable.
Conducting Factor AnalysisCalculate Factor Scores
19 19Results of Principal Components
-
8/14/2019 Class Factor Analysis
19/21
19-19Results of Principal ComponentsAnalysis
The lower left triangle contains the reproducedcorrelation matrix; the diagonal, the communalities;the upper right triangle, the residuals between theobserved correlations and the reproducedcorrelations.
The lower left triangle contains the reproducedcorrelation matrix; the diagonal, the communalities;the upper right triangle, the residuals between theobserved correlations and the reproducedcorrelations.
Fact
Variables V1 19 20C d ti F t A l i
-
8/14/2019 Class Factor Analysis
20/21
19-20
Selection of S/S variable, involves singling out someof the original variables for use in subsequent analysis& interpret the result in terms of original variablesrather than factor scores.
By examining the factor matrix, one could select foreach factor the variable with the highest loading onthat factor. That variable could then be used as asurrogate variable for the associated factor.
However, the choice is not as easy if two or morevariables have similarly high loadings. In such a case,
the choice between these variables should be basedon theoretical and measurement considerations.
Conducting Factor AnalysisSelect Surrogate Variables
19-21
C d ti F t A l i
-
8/14/2019 Class Factor Analysis
21/21
19-21
If there are many large residuals, the factormodel dies not provide a good fit to the dataand the model should be reconsidered .
Conducting Factor AnalysisDetermine the Model Fit