Exploratory Factor Analysis (EFA)

Mary Ann Coughlin

Springfield College

William Knight
Bowling Green State University

Welcome & Agenda

First of a three part series

Today Exploratory Factor Analysis

Next Confirmatory Factor Analysis

Final Structural Equation Modeling

Agenda:

Overview of EFA

Case Study Application with Senior Survey

Statistical Basis for EFA

Interpreting EFA output

Next Steps

1
Factor Analysis

Purpose:

Factor analysis explores the inter-
relationships among variables to discover if
those variables can be grouped into a
smaller set of underlying factors.

Applications of Factor Analysis

Three primary applications of Factor Analysis include:

Explore data for patterns. Often a researcher is unclear if
items or variables have a discernible patterns. Factor
Analysis can be done in an Exploratory fashion to reveal
patterns among the inter-relationships of the items.

Data Reduction. Factor analysis can be used to reduce a

large number of variables into a smaller and more
manageable number of factors. Factor analysis can create
factor scores for each subject that represents these higher
order variables.

Confirm hypothesis of factor structure. In measurement

research when a researcher wishes to validate a scale with
a given or hypothesized factor structure, Confirmatory
Factor Analysis is used.

2
Case Study

Exploratory Factor Analysis

Case Study
Exploratory Factor Analysis

Data

Annually, the Office of Institutional Research conducts a
survey of graduating seniors. This instrument contains a
bank of 27-items that attempt to measure various aspects of
outcomes of an undergraduate education.

Purpose:

The researcher wishes to determine if these 27-items can
be organized or grouped into a smaller set of underlying
factors.

These 27-items are just too detailed for presentation
purposes.

Analysis:

Exploratory Factor Analysis

3
SPSS Analysis

4
Assumptions of Factor Analysis

Assumes that the observed variables are a linear
combination of some underlying of hypothetical or
unobservable factors.

Some of the factors are assumed to be common to
two or more variables and some are assumed to be
unique to each variable.

The factors or unobserved variables are assumed to
independent of one another.

All variables in a factor analysis must consist of at
least an ordinal scale. Nominal data are not
appropriate for factor analysis.

Basic Assumptions

Two statistics on the SPSS output allow you to look at

some of the basic assumptions.

Kaiser-Meyer-Olkin Measure of Sampling Adequacy, and

Bartlett's Test of Sphericity

Kaiser-Meyer-Olkin Measure of Sampling Adequacy

generally indicates whether or not the variables are
able to be grouped into a smaller set of underlying
factors.

High values (close to 1.0) generally indicate that a factor
analysis may be useful with your data.

If the value is less than .50, the results of the factor analysis
probably won't be very useful.

Let us look at ours for our Factor analysis.

10

5
SPSS Output

KMO and Bartlett's Test

Kaiser-Meyer-Olkin Measure of Sampling
Adequacy. .920

Bartlett's Test of Approx. Chi-Square 4992.010

Sphericity df 351
Sig. .000

So clearly our data support the use of factor

analysis and suggest that the data may be
grouped into a smaller set of underlying
factors.

What does Bartlett’s Test of Sphericity
explore?
11

SPSS Output
KMO and Bartlett's Test
Kaiser-Meyer-Olkin Measure of Sampling
Adequacy. .920

Bartlett's Test of Approx. Chi-Square 4992.010

Sphericity df 351
Sig. .000

Bartlett’s Test of Sphericity compares your correlation

matrix to an identity matrix

An identity matrix is a correlation matrix with 1.0 on the
principal diagonal and zeros in all other correlations.

So clearly you want your Bartlett value to be significant as you
are expecting relationships between your variables, if a factor
analysis is going to be appropriate!

Problem with Bartlett’s test occurs with large n’s as small
correlations tend to be statistically significant – so test may not
mean much!

12

6
Procedures for Factor Analysis

The first step in completing a factor analysis

is to measure the inter-relationships among
the items.

This step leads to the determination of the
appropriate number of factors.

When initially determining the appropriate
number of factors, one factor is identified for
each variable.

Obviously the researcher expects that the
number of useful factors will be substantially less.

However, if no relationship exists between the
variables then each variable would make it's own
unique factor.

13

Procedures for Factor Analysis

Multiple different statistical procedures exist by

which the number of appropriate number of
factors can be identified.

These procedures are called "Extraction Methods."

By default SPSS does what is called a Principal
Components Extraction method.

This Principal Components Method is simpler and until
more recently was considered the appropriate method for
Exploratory Factor Analysis.

Statisticians now advocate for a different extraction method
due to a flaw in the approach that Principal Components
utilizes for extraction.

Let us briefly explore this issue.

14

7
 Principal Components Extraction

In the Principal Components Analysis, inter-

item correlation coefficient Matrix is what is
analyzed to explore the inter-relationships
between the items to determine if the items
can be grouped together to represent a
smaller set of underlying factors.

15

Principal Components Extraction

Correlation "R" Matrix

The correlation (R) matrix Item 1 Item 2 Item 3 Item 4 Item 5 Item 6
represents the relationships Item 1 1.000 .678 .780 .240 .189 .065
between all items and is a Item 2 .678 1.000 .858 .380 .345 .188
complete matrix with 1.0 on Item 3 .780 .858 1.000 .189 .243 .058
Item 4 .240 .380 .189 1.000 .789 .834
the diagonal.
Item 5 .189 .345 .243 .789 1.000 .657

The 1.0 indicates the perfect
Item 6 .065 .188 .058 .834 .657 1.000
relationship the variable has with
itself.

The upper and lower elements of
the matrix are mirror images.

On the right is a sample
correlation matrix representing
the relationship between six
items on an attitude scale.

This matrix is much easier to
interpret than the one for our
analysis, which would be a 27 X
27 matrix!!
16

8
R Matrix

This R matrix reports:

strong positive relationships among items I, 2, and 3,
as well as strong positive relationships among items
4, 5, and 6;

in addition, items 1, 2, and 3 have weak relationships
with items 4, 5, and 6;

this pattern would be the first indication that a two-
factor model might be appropriate for this data.
C o rrelatio n "R" Matrix
Item 1 Item 2 Item 3 Item 4 Item 5 Item 6
Item 1 1.000 .678 .780 .240 .189 .065
Item 2 .678 1.000 .858 .380 .345 .188
Item 3 .780 .858 1.000 .189 .243 .058
Item 4 .240 .380 .189 1.000 .789 .834
Item 5 .189 .345 .243 .789 1.000 .657
Item 6 .065 .188 .058 .834 .657 1.000
17

Principal Components Extraction

The problem with the Principal Components

Analysis is that correlation matrix has 1.0 on
the diagonal.

As a result the initial communalities among the
items are set or assumed to be 1.0.

So what are communalities and why is the initial

communalities in Principal Component Analysis
set to 1.0.

18

9
Principal Components Extraction

A communality (C) is the extent to which an

item correlates with all other items.

Thus in Principal Components Extraction method
when the initial communalities are set to 1.0, then
all of the variability of each item is accounted for in
the analysis.

Of course some of the variability is explained and
some is unexplained.

So in the Principal Components analysis with these

initial communalities set to 1.0, you are trying to find
both the common factor variance and the unique or
error variance.

19

Principal Components Extraction

In a two factor model, the initial communality of Item 1 can

be described by the following formula:

Item 1 = aFactor 1 + aFactor 2 + U item 1

Where

aFactor 1 + aFactor 2 represent the explained common

factor variance for the two factors, and

U item 1 represents the uniqueness of the item or error

variance.

The presence of the uniqueness in the model is what
makes the initial communalities add up to 1.0.

20

10
Principal Components Extraction

Statisticians have indicated that assuming that all of

the variability of the items whether explained or
unique can be accounted for in the analysis is
flawed and definitely should not be used in an
exploratory factor model.

Some researchers suggest Principal Axis Factoring as the

appropriate method for factor extraction using Exploratory
Factor Analysis.

21

Principal Axis Factoring

In Principal Axis factor extraction, the amount of variability each
item shares with all other items is determined and this value is
inserted into the R matrix replacing the 1.0 on the diagonals.

As a result, Principal Axis factoring is only analyzing common
factor variability; removing the uniqueness or unexplained
variability from the model.

In a two factor model, the initial communality of Item 1 can be
described by the following formula:

Item 1 = aFactor 1 + aFactor 2

Where:

aFactor 1 + aFactor 2 represent the explained common
factor variance for the two factors, and

the uniqueness of the item or error variance or U item 1 is

no longer in the formula.

22

11
 Principal Axis Factoring

Thus, when reviewing communalities,

As the values of the communalities decrease, then
the more unexplained variability or uniqueness exists
within that item.

As a result, lower communalities indicate that the
item does not add to the proposed factor structure.

In layperson's terms, an item with a lower
communality is a problem item, because the item is
not common to the proposed factor structure; rather
it is more unique or outside the factor structure.

Now let us look at our initial communalities from

our case study.
23

SPSS Analysis

Note:
1) These communalities are from a
Principal Components Analysis.
2) They are shown for illustrative purpose
to compare with the Principal Axis
method which is the recommended
procedure.
3) You should notice that:
a) The initial communalities are set to
1.0, due to the use of the correlation
matrix and the 1.0 on the diagonal.
b) Also the extracted communalities will
be higher using the Principal
Components analysis, due to the
inclusion of the uniqueness of each
item.

24

12
 SPSS Analysis
PAF Extraction

25

SPSS Analysis
Note:
1) These communalities are from a
Principal Axis Analysis, which is
the recommended procedure.

This item has a low extracted

communality.
Thus, the item does not add to
the proposed factor structure.
By itself a low communality
does not mean an item should
be dropped, but it does mean
that the item is problematic.

26

13
Eigenvalues

After the initial factor structure has been

extracted, the next step is to calculate
eigenvalues.

Remember, the initial factor structure starts with as
many factors as there are items in the analysis.

Eigenvalues represent the amount of variance

in the data that is explained by the factor with
which it is associated.

27

Eigenvalues

Common Characteristics of Eigenvalues

The factors are extracted in order by the amount
of variance that they explain. Therefore, the first
factor will have the highest eigenvalue, the
second the next highest, etc.

The first few factors generally explain the majority
of the variance with the last few explaining only a
very small proportion of variance.

28

14
Eigenvalues

Common Characteristics of Eigenvalues

At this stage in the analysis, the researcher must
determine the number of meaningful factors.

Determining the number of factors is not a
straightforward task, because the decision is
ultimately subjective.

Serveral criteria do exist for determining the
number of factors, but most of these are just
empirical guidelines.

Let us look at one such guideline –

Eigen One Rule or Kaiser-Guttman rule!

29

Eigenvalues

Eigen One Rule or Kaiser-Guttman rule!

This rule instructs you to keep only those factors
whose eigenvalues are greater than 1.0 and discard
the rest.

The logic of this rule is that by selecting 1.0 as
the criteria for retaining the factor, the variance
accounted for by the factor must be at least as
large as the variance of a single standardized
variable.

Remember, standardized variables (z values) have a
mean of 0 and a standard deviation of 1.0.

And if standard deviation is 1.0 – then variance must
also be 1.0.

30

15
SPSS Analysis

6 Factors exist
that explain a
variance
equivalent to at
least one
standardized
variable

31

Eigenvalues

Criticism of Eigen One Rule-

Many statistician criticize the eigen one rule as still being
somewhat arbitrary and based solely on the data.

These statistician feel that theory and logic should also be
considered in this process.

So as a result they suggest that the eigen one rule be used
solely as a guideline.

Before we proceed to discuss the factor matrix and how
eigen values are calculated, let us look at one more way
of determining the appropriate number of factors to
extract.

32

16
Scree Plot

A scree plot can be used to visually determine the
number of useful factors to be extracted.

A scree plot is a line graph with

Eigen values plotted on the Y or vertical axis, and

The factors are plotted on the horizontal or X-axis.

A scree plot should form the intersection of two lines

One line should be an initial steep line of useful factors and

The second line should be a gradual trailing line of factors
that should be eliminated.

The plot is called a 'Scree' Plot because it often looks
like a 'scree' slope, where rocks have fallen down and
accumulated on the side of a mountain, where the
scree refers to the debris that has fallen!

Let us look at our scree plot!

33

SPSS Scree Plot!

10

? Where do the Gradual

Initial 8 two lines Trailing
Steep intersect ? line!
line! e6
lu
a
v
n
e
ig 4
E

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
Factor Number

34

17
Interpretation of Scree Plot

To interpret our Scree plot we want to find the

point where the two lines intersect.

Keep all factors that fall on the initial steep line and

Discard the factors that are found in the gradual
trailing line or scree!

The problem with our data is that the

intersection between the two lines is not clear.

That intersection could occur anywhere between
factors 6 through 10.

So now what?

35

Interpretation of Extracted Factors

Well given that:

the application of the eigen one rule had us with 6
extracted factors, and

we do not have any theory that is driving the
creation of this instrument, and

that the scree plot could be interpreted as
supporting 6 meaningful factors,

it would make the most amount of sense to proceed with
the interpretation based upon the eigen one rule.

Let us go back and review our SPSS output

with the extracted factors.

36

18
SPSS Analysis

37

Extraction

Notice that all statistics produced up through the initial
eigenvalues were calculated for all items.

Remember, exploratory factor analysis starts with the
assumption that if the items can not be logically
grouped into a smaller set of underlying factors, then
each item would make a unique factor.

The eigenvalues represent the percent of explained
variance that is due to each factor.

Eigenvalues are calculated by squaring and then summing
the factor loadings for each item within each factor.

Thus to understand eigenvalues we must review the
factor matrix.

38

19
Factor Matrix

After the appropriate number of factors has

been determined, a factor matrix is
calculated. This matrix identifies the
relationship between the variables and the
factors.

In general, this matrix reads like a correlation
matrix and the factor loadings are similar in
structure and value to correlation coefficients.

Before, interpreting this initial factor matrix,
we must be aware of some general problems
with this initial matrix.

39

Three Problems with Initial Factor Matrix

Due to the fact that the first factor explains the most amount of
variance, most of the variables will have at least some
relationship with this first factor.

Thus, this factor becomes very generalized and difficult to interpret.

Due to first factor being a general factor, many variables may
load on more than one factor (double loading).

While this complexity, is not a problem statistically, it does add
needlessly to the complexity of the factor structure.

Many factors may be bi-polar. A bi-polar factor is one in which
both significant positive and negative loadings exist.

A negative loading is like a negative correlation coefficient.

Often negative loadings may be found which are due to the actual
values in the data.

Bipolar factors can create negative loadings which make no apparent
sense from the data.

40

20
Initial Factor Matrix

Let us take a look at our initial factor matrix.

It presents some of these problems.

As we interpret the factor matrix, remember
factor loadings are like correlation coefficients.

As a general guideline, factor loadings greater
than .40 indicate that an item is related or
associated with a given factor.

41

Factor Matrix

42

21
Factor Matrix

Before the rotation, process we have many items that

are loading on our first generalized factor.

In addition, we have a few items with double factor
loadings (i.e., items that load on more than one
factor).

Factor two is bipolar, with both positive and negative
factor loadings.

These problems are easier to identify if we sort the
matrix and suppress values that are low (< .30).

43

Factor Matrix

44

22
Rotation of the Factor Matrix

The factor matrix is rotated to address the

three common problems with the initial matrix.

Three common procedures exist for rotation:

Orthogonal –

Varimax

Quartimax,

Oblique

Each method varies in how the rotation is
accomplished.

45

Orthogonal Rotation -- Varimax

Orthogonal rotation is easiest of the three methods to describe
conceptually. However, it is also the most limited in terms of
it's application.

Orthogonal rotation has a restriction which states that the
factors may only be rotated in such a manner that the factors
are kept at right angles to each other.

This restriction follows the assumption that no association
exists between the factors. For many applications, the factors
are logically correlated, thus the limited application of this
procedure.

Graphically, the rotation procedures are illustrated as follows:

46

23
Orthogonal Rotation

Factor 1
Note:
+1.0

The number lines
represent the factor
loadings on two
.60
factors.

The right angle
Factor 2 restrictions
represents the fact no
-.80 -.40 .40 .80
relationship is
assumed to exist
- .60
between the two
factors.

-1.0

47

Orthogonal Rotation

Factor 1 Note:
1) X represents the variables
+1.0
and plots the factor loadings
X X for the two factors.
.60 X X 2) Some negative factor
X
X loadings appear for some
Factor 2 variables on factor 1.
3) The factors are now rotated
X X or shifted to improve the
-.80 -.40 .40 .80 relationships between the
X X
X variables and each factor,
X X
keeping the right angle
- .60 restriction.

-1.0

48

24
Orthogonal Rotation

Note:
1) The rotation increased the
Fac
tor strength of the relationship
+1
.0 1 between many of the variables
X X
X X
and the factors.
-.8
0 .60 X X 2) Also, the rotation eliminated the
negative factor loadings.
-.4
0

X X

.40 X X
X
X X F
act
-.
60 .80 or
2
-1.
0

49

Factor Transformation Matrix

The factor transformation matrix describes the

specific rotation applied to your factor solution.

This matrix is used to compute the rotated factor
matrix from the initial (unrotated) factor matrix.

For example:

50

25
 SPSS Analysis

Varimax Rotation

51

Factor Transformation Matrix

If the off-diagonal elements are close to zero,

the rotation was relatively small.

If the off-diagonal elements are large (greater
than ± .5), a larger rotation was applied.

Factor Transformation Matrix

Factor 1 2 3 4 5 6
1 .652 .398 .422 .331 .244 .263
2 .377 -.675 -.221 .093 .556 -.190
3 -.568 .035 .305 .625 .358 -.255
4 -.298 .217 -.267 -.257 .607 .598
5 .144 .574 -.556 .040 .183 -.553
6 .033 -.089 -.547 .651 -.319 .408
Extraction Method: Principal Axis Factoring.
Rotation Method: Varimax with Kaiser Normalization.

52

26
Next Step

Our next step would be to interpret the factor

matrix, which we ran with the values sorted by
size and suppressed values less than.30.

But before we do let us finish our discussion of
other rotation procedures.

53

Orthogonal -- Uncorrelated Model

Remember, EFA with varimax rotation is an

uncorrelated model.

This concept is maintained with orthogonal rotation,
as the factors remain at 90o angles, which assume
the factors are uncorrelated.

In addition, varimax rotation adds one additional
feature to orthogonal rotation as it maximizes the
amount of variance explained by each of the factors.

54

27
Other Rotation Procedures

Oblique Rotation:

Factors are rotated without maintaining the right
angle restrictions.

This allows the researcher to account for the
relationships that may logically exist between the
factors.

The closer the rotation angle is to 0, the higher the
correlation between the factors.

As the rotation angle approaches 90 degrees, the
relationship between the factors nears 0.

SPSS calls this Direct OBLIM!

55

Oblique Rotation

Fact Note:
or 1
+1
.0 1) The right angle restriction is no
longer held.
X X
2) In the current example, only a
.60 X X
X
X slight relationship is found
between the factor as the factors
-.80 X are still at about 80 degree
-.40 X
F
X X actor
angles.
.40 X 2
X X
.80

- .6
0

-1.
0

56

28
Other Rotation Procedures

Quartimax is another form of orthogonal

rotation.

It is used when you are expecting one
factor and

You are testing to see if there is one factor!

For our case study – we have run

Varimax let us look at the output and
interpret our factor analysis

57

Factor Matrix

We start by interpreting the factor matrix!

So you know, a factor matrix contains the
relationship between each item and each factor.

And the relationships are referred to as factor loadings.

We use the rotated factor matrix, not the initial

factor matrix, to interpret these relationships.

Because the rotated factor matrix address those three
common problems.

One set of procedures for interpreting the rotated

factor matrix is to use Thurstone's Simple Structure:

The term simple structure implies that items should be
related to only one factor.

These guidelines are helpful, if you are trying to create or
revise a scale.

58

29
Factor Matrix

Thurstone's Rules state:

Select items that relate strongly to the proposed factor (i.e.,
factor loadings of .40 or above).

Delete or drop items that are double loaded (i.e., .40 or above
on more than one factor).

Delete items that are unique or do not load on any factor (i.e.,
all factor loadings are below .40).

Delete items that load high on a factor that was not the
proposed factor.

Some statisticians acknowledge that factor complexity

is inevitable.

As a result, some researchers will maintain items with double
factor loadings as long as the items would logically belong to
both factors.

If greater than .05 percent difference exists between the two
factor loadings, the item can be considered to primarily
belonging to the factor with the higher factor loading.

59

Factor Matrix

60

30
Factor Matrix

To better understand this rotated matrix, we

can use the computer to organize our factor
matrix, by telling SPSS to sort the matrix and
suppress values that are less than .40.

You may want to consider suppressing values less
than .30, so that you can see what items are close
to loading on some factors.

61

Factor Matrix

62

31
Factor Matrix

So what have we found?

What do our factors represent?

63

Interpretation of the Factors

Analysis of our Rotated Factor matrix from the

exploratory factor analysis revealed a total of six
factors.

After discussion and consultation, these factors were
named:

Intellectual

Moral

Self-Development ~ Self-Awareness

Leadership

Technology

Humanities

64

32
Factor Matrix

65

Interpretation of the Factors

The naming of the factors is the responsibility of the

researcher.

When naming the factors you may consider items that
have higher factor loadings as being more
representative of the factor than items with lower factor
loadings.

Naming of the factors is an important and difficult stage
in the analysis.

In some cases, the researcher has some
predetermined structure that is used in this phase.

However, if a proposed factor structure does exist, the researcher
would eventually want to complete a confirmatory factor analysis.

When naming factors, do not do so in isolation.

Be sure that you do so in the context of your research and
organization

66

33
Making Connections

Now that we have interpreted and

named our factors, let us go back and
make a few connections between the
factor matrix, factor loadings,
communalities, and eigenvalues.

Let us start with the communalities

Do you remember what a communality
is?

67

SPSS Analysis A communality is

the proportion of a
variable's variance
explained by the
factor structure

So what is the
difference between
initial and extraction
communality and why
are they different?

68

34
Initial Communality

Initial Communality is equal to:

R21.23456789 – k

You can easily calculate this with the SPSS

REGRESSION program

entering as the dependent variable the variable that you are
trying to calculate the communality for (e.g., variable 1) and

Then enter all the remaining variables in the factor analysis
as the predictor variables

Doesn’t all this make sense given our definition of a

communality?

69

SPSS Regression Analysis

Variables Entered/Removedb

Variables
Model Variables Entered Removed Method
1 Understand the role of science and technology,
Develop feminist awarenenss, Read or speack
a foreign language, Form close friendships,
Here I have entered all
Gain in-depth knowlegde of a field, Use
computers for basic tasks (word processing),
Place current problems in historical my 26 other variables as
prospective, Lead and supervise tasks and
groups of people, Think analytically and
logically, Use computers for complex tasks the predictor variables!
(graphing), Acquire broad knowledge in the Arts
and Sciences, Appreciate art, literature, music,
drama, Relate well to people of different races,
nations, Communicate well orally, Acquire new . Enter
skills and knowledge on my own, Plan and
execute complex projects, Identify moral and
ethical issues, Function independently without
supervision, Function effectively as a member
of a team, Evaluate and choose between
alternative courses, Understand myself, my
abilities, interests, Formulate creative / original
ideas and solutions, Synthesize and integrate
ideas and information, Develop awareness of
social problems, Establish a course of action to
accomplish goals, a
Develop self-esteem
/self-confidence
a. All requested variables entered.
b. Dependent Variable: Write effectively
Here I have entered my first
variable write effectively as
my dependent variable,
70

35
 SPSS Regression Analysis
Model Summary

Adjusted R Std. Error of the

Model R R Square Square Estimate
1 .605a .366 .334 .530
a. Predictors: (Constant), Understand the role of science and Notice that the R2 for
technology, Develop feminist awarenenss, Read or speack a foreign
language, Form close friendships, Gain in-depth knowlegde of a this analysis is equal to
field, Use computers for basic tasks (word processing), Place
current problems in historical prospective, Lead and supervise tasks
and groups of people, Think analytically and logically, Use
the communality for the
computers for complex tasks (graphing), Acquire broad knowledge in
the Arts and Sciences, Appreciate art, literature, music, drama, first item in the factor
Relate well to people of different races, nations, Communicate well
orally, Acquire new skills and knowledge on my own, Plan and
execute complex projects, Identify moral and ethical issues, Function
analysis – write
independently without supervision, Function effectively as a member
of a team, Evaluate and choose between alternative courses, effectively!
Understand myself, my abilities, interests, Formulate creative /
original ideas and solutions, Synthesize and integrate ideas and
information, Develop awareness of social problems, Establish a
course of action to accomplish goals, Develop self-esteem
/self-confidence

71

SPSS Analysis
See they match!

So, now how do we

calculate extracted
communalities, and
and why are they
different?

72

36
Communalities

We can manipulate the factor matrices, both

initial and rotated to further illustrate the
calculations of the communalities.

If we take either the initial factor or the rotated
factor matrix and square the factor loadings;

the extracted communalities for each item are the sum of
the squared factor loadings across the factors or rows.

Why doesn’t it matter whether I use the initial or
rotated factor matrix?

Because in either matrix all the common variance that the
item shares with the factor structure is the same, it is just
“rotated” or distributed differently across the factors in the
rotated factor matrix!

73

Initial Factor Matrix

74

37
Communalities

Factor
1 SQ -1 2 SQ -2 3 SQ -3 4 SQ -4 5 SQ - 5 6 SQ -6 Communalities
Write effectively .5228 .2733 -.0335 .0011 -.2744 .0753 -.0476 .0023 .0938 .0088 .0200 .0004 .3612
Communicate well orally .5814 .3380 .0421 .0018 -.0763 .0058 -.1123 .0126 -.0910 .0083 .0512 .0026 .3691
Acquire new skills and .5579 .3113 .1744 .0304 -.2315 .0536 -.0736 .0054 -.0147 .0002 .0235 .0006 .4015
Think analytically and logically .5415 .2932 .1927 .0371 -.2793 .0780 -.1290 .0166 .1639 .0269 -.0864 .0075 .4593
Formulate creative / original ideas .6325 .4000 .1175 .0138 -.2018 .0407 -.1519 .0231 .0382 .0015 .0987 .0098 .4889
Evaluate and choose .6307 .3978 .0982 .0096 .0465 .0022 -.1634 .0267 .0671 .0045 .1621 .0263 .4670
Lead and supervise .5196 .2699 .1063 .0113 .4529 .2051 -.1507 .0227 .0012 .0000 .3380 .1142 .6233
Relate well to people of different .5171 .2674 -.1464 .0214 .2766 .0765 .0084 .0001 .0271 .0007 -.0343 .0012 .3673
Function effectively as a member of a team .5244 .2750 .0646 .0042 .4101 .1682 -.0392 .0015 .0620 .0038 .1826 .0334 .4861
Use computers for basic tasks .3788 .1435 .1350 .0182 .2078 .0432 .1711 .0293 -.0037 .0000 -.1772 .0314 .2656
Use computers for complex tasks .2057 .0423 .4090 .1672 .2383 .0568 .2928 .0857 .1362 .0186 -.1119 .0125 .3832
Place current problems in .4787 .2291 -.2436 .0594 -.1143 .0131 .0308 .0010 .1562 .0244 .1582 .0250 .3519
Identify moral and ethical issues .5927 .3513 -.2461 .0605 .0254 .0006 .0741 .0055 .2150 .0462 .0100 .0001 .4643
Understand myself, my abilities, interests .6342 .4023 -.2287 .0523 .0305 .0009 -.1656 .0274 -.1891 .0358 -.1279 .0164 .5351
Function independently without supervision .6201 .3846 .0436 .0019 .0482 .0023 -.0414 .0017 -.2092 .0438 -.1101 .0121 .4464
Gain in-depth knowlegde of a field .4336 .1880 .1694 .0287 -.1718 .0295 .0200 .0004 .0277 .0008 -.0131 .0002 .2476
Plan and execute complex projects .5695 .3244 .2557 .0654 -.1332 .0177 .0168 .0003 -.0145 .0002 .0359 .0013 .4093
Read or speak a foreign language .2299 .0529 -.0971 .0094 -.1040 .0108 .2365 .0559 -.2020 .0408 .1069 .0114 .1813
Appreciate art, literature, music, drama .4621 .2135 -.2769 .0767 -.1792 .0321 .2963 .0878 -.2758 .0761 .2462 .0606 .5468
Acquire broad knowledge .4443 .1974 .0836 .0070 -.1043 .0109 .3322 .1104 -.1012 .0102 .0758 .0057 .3416
Develop feminist awarenenss .3796 .1441 -.4281 .1833 -.0193 .0004 .1229 .0151 .1584 .0251 -.0355 .0013 .3692
Develop awareness of social problems .5736 .3290 -.4395 .1931 .0748 .0056 .1468 .0216 .2813 .0792 -.1206 .0145 .6430
Develop self-esteem /self-confidence .7160 .5127 -.1546 .0239 .0628 .0039 -.0748 .0056 -.1341 .0180 -.1700 .0289 .5930
Form close friendships .4726 .2234 -.1496 .0224 .2055 .0422 -.0921 .0085 -.2318 .0537 -.2191 .0480 .3982
Establish a course of action .7169 .5140 .0826 .0068 .0267 .0007 -.0755 .0057 -.0470 .0022 -.0866 .0075 .5369
Synthesize and integrate ideas .6611 .4371 .1493 .0223 -.1332 .0178 -.0485 .0024 .0827 .0068 -.0807 .0065 .4928
Understand the role of science .5087 .2588 .4309 .1857 .0601 .0036 .2694 .0726 .0171 .0003 -.0599 .0036 .5246
Eigenvalues -- Extracted 7.7742 1.3151 .9976 .6478 .5369 .4829

75

SPSS Analysis
See they match!

So, why are they

different?

76

38
The Difference!

The initial communalities represent the amount of

variance of the variable that is explained by all the
variables in the model

Hence it can be replicated by the R2 value, and

Since these are initial values prior to extraction, it represents
the amount of variance of the variable that is shared by all
other variables.

Remember, the initial assumption in EFA is that we will have
as many factors as we have variables.

The extracted communalities represent the amount of

variance of the variable that is explained by the factor
structure.

Thus it is calculated by squaring and summing the factor
loadings.

77

Eigenvalues

We can also manipulate the initial and rotated

factor matrix and further illustrate the
calculations of the eigenvalues.

If we take the initial factor matrix and square the
factor loadings:

The extracted eigenvalues are the sum of the squared
factor loadings across the items or columns.

78

39
 Eigenvalues

Factor
1 SQ -1 2 SQ -2 3 SQ -3 4 SQ -4 5 SQ - 5 6 SQ -6 Communalities
Write effectively .5228 .2733 -.0335 .0011 -.2744 .0753 -.0476 .0023 .0938 .0088 .0200 .0004 .3612
Communicate well orally .5814 .3380 .0421 .0018 -.0763 .0058 -.1123 .0126 -.0910 .0083 .0512 .0026 .3691
Acquire new skills and .5579 .3113 .1744 .0304 -.2315 .0536 -.0736 .0054 -.0147 .0002 .0235 .0006 .4015
Think analytically and logically .5415 .2932 .1927 .0371 -.2793 .0780 -.1290 .0166 .1639 .0269 -.0864 .0075 .4593
Formulate creative / original ideas .6325 .4000 .1175 .0138 -.2018 .0407 -.1519 .0231 .0382 .0015 .0987 .0098 .4889
Evaluate and choose .6307 .3978 .0982 .0096 .0465 .0022 -.1634 .0267 .0671 .0045 .1621 .0263 .4670
Lead and supervise .5196 .2699 .1063 .0113 .4529 .2051 -.1507 .0227 .0012 .0000 .3380 .1142 .6233
Relate well to people of different .5171 .2674 -.1464 .0214 .2766 .0765 .0084 .0001 .0271 .0007 -.0343 .0012 .3673
Function effectively as a member of a team .5244 .2750 .0646 .0042 .4101 .1682 -.0392 .0015 .0620 .0038 .1826 .0334 .4861
Use computers for basic tasks .3788 .1435 .1350 .0182 .2078 .0432 .1711 .0293 -.0037 .0000 -.1772 .0314 .2656
Use computers for complex tasks .2057 .0423 .4090 .1672 .2383 .0568 .2928 .0857 .1362 .0186 -.1119 .0125 .3832
Place current problems in .4787 .2291 -.2436 .0594 -.1143 .0131 .0308 .0010 .1562 .0244 .1582 .0250 .3519
Identify moral and ethical issues .5927 .3513 -.2461 .0605 .0254 .0006 .0741 .0055 .2150 .0462 .0100 .0001 .4643
Understand myself, my abilities, interests .6342 .4023 -.2287 .0523 .0305 .0009 -.1656 .0274 -.1891 .0358 -.1279 .0164 .5351
Function independently without supervision .6201 .3846 .0436 .0019 .0482 .0023 -.0414 .0017 -.2092 .0438 -.1101 .0121 .4464
Gain in-depth knowlegde of a field .4336 .1880 .1694 .0287 -.1718 .0295 .0200 .0004 .0277 .0008 -.0131 .0002 .2476
Plan and execute complex projects .5695 .3244 .2557 .0654 -.1332 .0177 .0168 .0003 -.0145 .0002 .0359 .0013 .4093
Read or speak a foreign language .2299 .0529 -.0971 .0094 -.1040 .0108 .2365 .0559 -.2020 .0408 .1069 .0114 .1813
Appreciate art, literature, music, drama .4621 .2135 -.2769 .0767 -.1792 .0321 .2963 .0878 -.2758 .0761 .2462 .0606 .5468
Acquire broad knowledge .4443 .1974 .0836 .0070 -.1043 .0109 .3322 .1104 -.1012 .0102 .0758 .0057 .3416
Develop feminist awarenenss .3796 .1441 -.4281 .1833 -.0193 .0004 .1229 .0151 .1584 .0251 -.0355 .0013 .3692
Develop awareness of social problems .5736 .3290 -.4395 .1931 .0748 .0056 .1468 .0216 .2813 .0792 -.1206 .0145 .6430
Develop self-esteem /self-confidence .7160 .5127 -.1546 .0239 .0628 .0039 -.0748 .0056 -.1341 .0180 -.1700 .0289 .5930
Form close friendships .4726 .2234 -.1496 .0224 .2055 .0422 -.0921 .0085 -.2318 .0537 -.2191 .0480 .3982
Establish a course of action .7169 .5140 .0826 .0068 .0267 .0007 -.0755 .0057 -.0470 .0022 -.0866 .0075 .5369
Synthesize and integrate ideas .6611 .4371 .1493 .0223 -.1332 .0178 -.0485 .0024 .0827 .0068 -.0807 .0065 .4928
Understand the role of science .5087 .2588 .4309 .1857 .0601 .0036 .2694 .0726 .0171 .0003 -.0599 .0036 .5246
Eigenvalues -- Extracted 7.7742 1.3151 .9976 .6478 .5369 .4829

79

Eigenvalues

We can also manipulate the initial and rotated

factor matrix and further illustrate the
calculations of the eigenvalues.

This time we will take the final (rotated) factor
matrix and square the factor loadings:

The rotated eigenvalues are the sum of the squared factor
loadings across the items or columns.

80

40
 Rotated Factor Matrix
Rotated Factor Matrixa

Factor
1 2 3 4 5 6
Write effectively .512 .263 .094 .027 -.007 .142
Communicate well orally .460 .119 .265 .207 .037 .168
Acquire new skills and knowledge on my own .581 .070 .141 .090 .095 .146
Think analytically and logically .643 .150 .091 .006 .118 -.026
Formulate creative / original ideas and solutions .625 .146 .145 .199 .031 .124
Evaluate and choose between alternative courses .485 .175 .176 .397 .087 .067
Lead and supervise tasks and groups of people .178 .089 .189 .724 .149 .048
Relate well to people of different races, nations .125 .335 .337 .307 .165 .070
Function effectively as a member of a team .160 .190 .208 .567 .240 .038
Use computers for basic tasks (word processing) .122 .118 .247 .108 .402 .053
Use computers for complex tasks (graphing) .082 -.034 -.024 .112 .601 -.030
Place current problems in historical prospective .303 .433 .040 .166 -.063 .198
Identify moral and ethical issues .288 .542 .168 .185 .098 .126
Understand myself, my abilities, interests .328 .275 .547 .159 -.056 .156
Function independently without supervision .372 .100 .455 .170 .165 .189
Gain in-depth knowlegde of a field .442 .074 .080 .039 .160 .117
Plan and execute complex projects .537 .042 .127 .148 .230 .168
Read or speack a foreign language .076 .079 .078 .003 .038 .402
Appreciate art, literature, music, drama .178 .248 .141 .088 -.054 .650
Acquire broad knowledge in the Arts and Sciences .269 .124 .064 .050 .277 .413
Develop feminist awarenenss .082 .560 .148 .025 -.037 .158
Develop awareness of social problems .158 .732 .233 .091 .102 .099
Develop self-esteem /self-confidence .370 .313 .543 .165 .096 .162
Form close friendships .122 .163 .569 .143 .077 .084
Establish a course of action to accomplish goals .496 .195 .386 .223 .204 .112
Synthesize and integrate ideas and information .587 .202 .217 .113 .208 .072
Understand the role of science and technology .380 -.013 .089 .138 .571 .164
Extraction Method: Principal Axis Factoring.
Rotation Method: Varimax with Kaiser Normalization.
a. Rotation converged in 7 iterations.

81

Eigenvalues

Factor
1 SQ -1 2 SQ -2 3 SQ -3 4 SQ -4 5 SQ - 5 6 SQ -6 C o mmunalities
W rite effectively .5122 .2623 .2628 .0691 .0941 .0089 .0275 .0008 -.0074 .0001 .1417 .0201 .3612
C o mmunicate well o rally .4602 .2118 .1192 .0142 .2655 .0705 .2072 .0429 .0369 .0014 .1685 .0284 .3691
A cquire new skills and .5814 .3380 .0698 .0049 .1413 .0200 .0899 .0081 .0953 .0091 .1464 .0214 .4015
Think analytically and lo gically .6433 .4138 .1496 .0224 .0913 .0083 .0060 .0000 .1184 .0140 -.0259 .0007 .4593
Formulate creative / o riginal ideas .6251 .3908 .1455 .0212 .1448 .0210 .1990 .0396 .0307 .0009 .1239 .0154 .4889
Evaluate and choo se .4854 .2356 .1750 .0306 .1764 .0311 .3971 .1577 .0865 .0075 .0668 .0045 .4670
Lead and supervise .1779 .0316 .0888 .0079 .1887 .0356 .7236 .5236 .1490 .0222 .0482 .0023 .6233
Relate w ell to people of different .1250 .0156 .3348 .1121 .3367 .1134 .3069 .0942 .1648 .0272 .0696 .0048 .3673
Function effectively as a member of a team .1600 .0256 .1903 .0362 .2084 .0434 .5672 .3218 .2400 .0576 .0381 .0015 .4861
Use co mputers for basic tasks .1224 .0150 .1178 .0139 .2468 .0609 .1083 .0117 .4015 .1612 .0533 .0028 .2656
Use co mputers for complex tasks .0816 .0067 -.0339 .0012 -.0236 .0006 .1124 .0126 .6011 .3613 -.0299 .0009 .3832
Place current pro blems in .3035 .0921 .4332 .1877 .0397 .0016 .1657 .0274 -.0626 .0039 .1981 .0392 .3519
Identify mo ral and ethical issues .2882 .0831 .5415 .2933 .1678 .0282 .1853 .0343 .0981 .0096 .1259 .0159 .4643
Understand myself, my abilities, interests .3277 .1074 .2747 .0755 .5471 .2994 .1595 .0254 -.0557 .0031 .1559 .0243 .5351
Function independently witho ut supervision .3718 .1382 .0998 .0100 .4545 .2066 .1700 .0289 .1645 .0271 .1887 .0356 .4464
G ain in-depth kno wlegde of a field .4416 .1950 .0736 .0054 .0796 .0063 .0394 .0016 .1598 .0255 .1171 .0137 .2476
Plan and execute co mplex pro jects .5373 .2887 .0416 .0017 .1272 .0162 .1476 .0218 .2295 .0527 .1680 .0282 .4093
Read o r speak a foreign language .0762 .0058 .0791 .0063 .0776 .0060 .0029 .0000 .0375 .0014 .4022 .1618 .1813
A ppreciate art, literature, music, drama .1785 .0319 .2483 .0616 .1414 .0200 .0884 .0078 -.0542 .0029 .6500 .4225 .5468
A cquire broad kno wledge .2691 .0724 .1239 .0154 .0635 .0040 .0497 .0025 .2766 .0765 .4133 .1708 .3416
Develo p feminist aw arenenss .0818 .0067 .5601 .3137 .1479 .0219 .0255 .0006 -.0372 .0014 .1577 .0249 .3692
Develo p aw areness o f so cial problems .1583 .0251 .7317 .5354 .2328 .0542 .0907 .0082 .1016 .0103 .0987 .0097 .6430
Develo p self-esteem /self-co nfidence .3701 .1369 .3134 .0982 .5434 .2952 .1650 .0272 .0956 .0091 .1619 .0262 .5930
Form clo se friendships .1218 .0148 .1628 .0265 .5687 .3235 .1426 .0203 .0773 .0060 .0841 .0071 .3982
Establish a co urse of actio n .4961 .2461 .1949 .0380 .3863 .1493 .2228 .0496 .2036 .0414 .1118 .0125 .5369
Synthesize and integrate ideas .5865 .3440 .2019 .0408 .2166 .0469 .1127 .0127 .2079 .0432 .0720 .0052 .4928
Understand the ro le of science .3800 .1444 -.0126 .0002 .0891 .0079 .1385 .0192 .5709 .3259 .1642 .0270 .5246
Eigenvalues -- Rotated 3.8796 2.0431 1.9009 1.5008 1.3027 1.1274

82

41
SPSS Analysis
Total Variance Explained

Initial Eigenvalues Extraction Sums of Squared Loadings Rotation Sums of Squared Loadings
Factor Total % of Variance Cumulative % Total % of Variance Cumulative % Total % of Variance Cumulative %
1 8.311 30.783 30.783 7.774 28.793 28.793 3.880 14.369 14.369
2 1.859 6.885 37.667 1.315 4.871 33.664 2.043 7.568 21.936
3 1.556 5.762 43.430 .998 3.695 37.359 1.901 7.040 28.976
4 1.252 4.638 48.067 .648 2.399 39.758 1.501 5.558 34.535
5 1.095 4.054 52.121 .537 1.989 41.746 1.303 4.825 39.359
6 1.007 3.731 55.852 .483 1.788 43.535 1.127 4.176 43.535
7 .977 3.619 59.471
8 .905 3.352 62.823
9 .795 2.946 65.769
10 .769 2.850 68.618
11 .718 2.661 71.279
12 .686 2.540 73.819
13 .668 2.475 76.293
14 .632 2.342 78.636
15 .580 2.147 80.782
16 .567 2.098 82.881
17 .537 1.987 84.868
18 .518 1.920 86.788
19 .468 1.734 88.521
20 .453 1.679 90.201
21 .437 1.618 91.819
22 .425 1.575 93.393
23 .403 1.491 94.884
24 .376 1.392 96.276
25 .372 1.376 97.652
26 .333 1.233 98.885
27 .301 1.115 100.000
Extraction Method: Principal Axis Factoring.

83

Next Steps

How do we use the factors to analyze for

differences across individuals or groups?

We need scores that can be used to represent
the factors.

84

42
Factor Scores

Factor scores are like predicted scores for each

individual score for each factor.

Regression factor scores are calculated as:

the case's standardized score (Z score) on each variable,
multiplies by the corresponding factor loading of the variables
for the given factor, and sums these products.

These regression factor scores will have a mean of 0, but a
variance equal to the squared multiple correlations between the
estimated factor scores and the true factor values.

These scores can be correlated with each other even when the
factors are orthogonal

As a result these regression scores can be difficult to interpret,
as they will go beyond the normal range of standardized scores
(-3.0 -- +3.0).

85

Other Methods for Calculating Factor Scores

Other methods exist for calculating factor scores:

The Andersen-Rubin method creates a factor score that is
based upon a standardized scale.

Meaning that they will have a mean of zero and as standard
deviation of 1.0.

Making these scores easier to interpret.

Some researcher still prefer to use factor scores that are

based upon the mean or sum of the items in the factor.

These scores are generally easier to interpret as they are based
upon the scale of the variables.

However, these scores do not take into consideration the fact that
some of variables contribute more to explaining the variance of
each of the factors (i.e., they have higher factor loadings).

Of course the advantage of factor scores is that they allow

the researcher to compare differences between various
subgroups (e.g., gender) on the factor scores.

86

43
 SPSS Analysis

Factor Scores

87

Review and Next Steps:

So far, we have --

Provided an Overview of EFA

Completed a Case Study Application with Senior
Survey

Discussed the Statistical Basis for EFA

Interpreted EFA output

Next steps –

Confirmatory Factor Analysis

Model testing

Remember , today was the first of a three part
series

Next Confirmatory Factor Analysis

Final Structural Equation Modeling

88

44
For More Information

In case you missed it: recorded version and slides

available at www.spss.com/airseries

Product questions? Call SPSS at 1-800-543-2185 or
sales@spss.com

Please fill out the post event survey

89

Questions!

45