Professional Documents
Culture Documents
Factor Analysis 2
Factor Analysis 2
Factor Analysis
A factor is a weighted sum of the variables The goal is to summarize the information in a larger number of correlated variables into a smaller number of factors that are not correlated with each other.
In contrast to Regression, there is no dependent variable. We just look at the correlations between variables to summarize.
Factor Analysis
2.3
C
2.2
Y
2.1
A
2.0 0.0 0.1 0.2 0.3 0.4
D
X
0.5 0.6 0.7 0.8 0.9 1.0
Factor Analysis
Graphical Intuition: Factor Analysis will not work when variables are uncorrelated
Figure 2
0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.1 0.2 0.3 0.4
0.5
0.6
0.7
0.8
0.9
1.0
Factor Analysis
Applications
Useful in constructing perceptual maps of products that are useful in positioning studies
Factor Analysis
VW Golf
Economy
1 0.5
Dodge Neon
Camry
-1.5 -1
Taurus
0 -0.5 -0.5 -1 -1.5 Fashion
Factor Analysis
0.5
1.5
Factor Analysis
Variables available
GPA GMAT score Which variables Scholarships, fellowships won do you believe Evidence of Communications skills correlate with Prior Job Experience intelligence and Organizational Experience Other extra curricular achievements teamwork and
leadership skills?
Factor Analysis
Data
Applicant 1 2 3
GPA 3.7
GMAT 680
Org. skills 3
Extracur ricular 2
20
Factor Analysis
Quick and dirty sense of the data: Looking at the correlation matrix
GMAT Fellowship 0.97 1.00 0.99 0.55 0.27 0.16 0.12 0.96 0.99 1.00 0.47 0.19 0.07 0.05
Comm Job Ex 0.43 0.55 0.47 1.00 0.82 0.79 0.69 0.05 0.27 0.19 0.82 1.00 0.99 0.98
Even if data is not as neatly correlated as here Factor analysis will be helpful
Factor Analysis
10
PCA uses the correlation matrix of the data and constructs factors
Factors If there are n variables we will have n factors First factor will explain most variance, second next, and so on Variance Explained by Factors With standardized variables each variable has a variance of 1, so the total variance in n variables is n Each factor will have an associated eigen-value which is the amount of variance explained by that factor
Factor Analysis
11
Component 1 2 3 4 5 6 7
12
Factor Analysis
13
Second Step: Do Factor Analysis with number of factors selected from Step 1
Use factor loadings to interpret factors If it is not interpretable use rotation options until we get something that can be interpreted
Factor Analysis
14
Why not Unrotated Factor Loadings? Variables correlation with the factors
Unrotated Factor Loadings and Communalities
Component Matrixa Component 1 2 .891 -.388 .766 -.586 .777 -.552 .883 .052 .683 .662 .518 .730 .493 .705
15
Factor Analysis
16
Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. a. Rotation converged in 3 iterations.
Factor Analysis
17
Factor Analysis
18
Naming Factors
Factor Analysis
19
Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. Component Scores.
Intelligence=0.293 GMAT + 0.315 GPA + 0.309 Fellowships + 0.181 Communications - 0.015 Job Ex - 0.068 Organizational Skills - 0.068 ExtraCurricular Leadership= - 0.006 GMAT - 0.097 GPA - 0.083 Fellowships + 0.153 Communications + 0.344 Job Ex + 0.343 Organizational Skills + 0.331 ExtraCurricular
Factor Analysis
20
Successful applicants
F2Score
No Good
Sure rejects
-1
-2
-2
-1
0 F1Score
Factor Analysis
21
Do Factor Analysis In SPSS select Analyze>Data Reduction>Factor Select Extraction, select Principle Component Analysis
Select the variables you want to factor analyze in Variables box Select Correlation as the data that will be analyzed; this will mean that the data will be standardized and therefore each variable will have equal effect. Ask for Scree Plot (using Graphs button) which graphs the amount of variance explained by each factor
Factor Analysis
22
Do Factor Analysis
Number of Factors to extract should be from Step 1 Try None rotation for a start (else try Varimax or others if it doesnt work) In Graphs: select loading plot and score plot In Storage: in the scores box store the factor scores by selecting 2 variables
Factor Analysis
23