SPSS Factor Analysis for Beginners 1 - Basic Ideas

Summary

In this SPSS tutorial, Ruben introduces the concept of factor analysis through a case study on client satisfaction with the Dutch unemployment agency. The agency uses 16 survey questions to assess their performance, with responses scored on a 7-point scale. Ruben explains the limitations of calculating mean scores and introduces factor analysis as a method to identify underlying factors from the survey data. He discusses exploratory factor analysis (EFA) in SPSS, which aims to determine the number of factors, which items measure them, and what real-world characteristics they represent. The tutorial also covers the use of eigenvalues to assess factor quality and the interpretation of factor loadings.

Takeaways

• 📊 **Factor Analysis Basics**: The video introduces the basic concept of factor analysis, which is a statistical method used to describe variability among observed variables in terms of fewer unobserved variables called factors.
• 🇳🇱 **Case Study Context**: The case study revolves around a Dutch unemployment agency aiming to assess client satisfaction through a survey, highlighting the practical application of factor analysis.
• 📝 **Survey Design**: The agency formulated 16 survey questions to measure client satisfaction, each scored on a 7-point Likert scale, demonstrating a common research methodology.
• 🤔 **Philosophical Questioning**: The video raises questions about whether client satisfaction is made up of separate aspects and if some statements might measure the same underlying sentiments.
• 📊 **Factor Analysis Diagram**: A diagram is used to illustrate how nine out of 16 statements might measure only three underlying factors, simplifying the data into more manageable components.
• 🔍 **Exploratory vs Confirmatory Factor Analysis**: The tutorial differentiates between exploratory factor analysis (EFA) and confirmatory factor analysis (CFA), with EFA being more about discovering the factor structure within the data.
• 🔢 **Correlation Analysis**: The video explains how correlations among variables can indicate the presence of underlying factors, with high correlations suggesting shared factors.
• 🟢 **Conditional Formatting**: A practical tip is given on using conditional formatting in Excel to visualize correlations, making it easier to identify patterns.
• 🧮 **Eigenvalues**: The concept of eigenvalues is introduced as a method to determine the number of factors to retain, with factors having eigenvalues over one considered significant.
• 📈 **Factor Loadings**: Factor loadings are mentioned as a way to interpret how strongly each item is related to a factor, which is crucial for understanding what each factor represents.
• 📊 **Dimension Reduction**: Factor analysis is positioned as a method of dimension reduction, simplifying data sets from multiple variables to a few underlying factors for easier analysis and reporting.

Q & A

What is the main purpose of the unemployment agency in the Netherlands mentioned in the script?

The main purpose of the unemployment agency in the Netherlands is twofold: first, to provide unemployed people with monthly income known as 'doll', and second, to assist unemployed individuals in finding employment.

What is the significance of the 16 survey questions formulated by the unemployment agency?

The 16 survey questions are significant as they were designed to measure client satisfaction, which helps the unemployment agency assess their performance in assisting and providing income to unemployed people.

What is the likert scale mentioned in the script, and how is it used?

The likert scale is a 7-point scale used to measure responses to statements, with each point representing a level of agreement or disagreement. In this case, it is used to score the 16 statements regarding client satisfaction.

Why might computing the mean scores for each statement not be the best method to evaluate performance according to the script?

Computing the mean scores might not be the best method because it doesn't account for the different aspects of client satisfaction that each statement might represent, nor does it consider the possibility that some statements might measure the same underlying sentiment.

What is the basic idea behind factor analysis as introduced in the script?

The basic idea behind factor analysis is to identify the underlying factors or dimensions that explain the correlations among a set of observed variables. In the context of the script, it is used to simplify the 16 satisfaction statements into a smaller number of factors.

What is the difference between exploratory factor analysis (EFA) and confirmatory factor analysis (CFA)?

Exploratory Factor Analysis (EFA) is used to identify the underlying structure of a set of variables, whereas Confirmatory Factor Analysis (CFA) is used to test whether observed data fit a proposed factor model.

What are the three basic questions EFA tries to answer according to the script?

The three basic questions EFA tries to answer are: 1) How many underlying factors are measured by the variables? 2) Which items measure which factors? 3) What real-world characteristics do the factors represent?

What is a Pieron correlation and how is it used in factor analysis?

A Pieron correlation is a numerical value between -1 and +1 that indicates the extent to which two quantitative variables are linearly related. In factor analysis, it is used to identify patterns of correlations among variables, which can suggest the underlying factors.

What is the significance of the eigenvalues in factor analysis?

Eigenvalues in factor analysis are used to determine the number of factors to retain. It is common practice to retain factors with eigenvalues greater than one, as they are considered to represent significant variance in the data.

How can factor loadings help in interpreting factors in factor analysis?

Factor loadings indicate the correlation between each item and the factors. They help in interpreting factors by showing which items have a stronger relationship with each factor, thus providing insight into what each factor represents.

What is the practical application of factor analysis mentioned in the script?

The practical application of factor analysis mentioned in the script is dimension reduction, where the original variables (statements) are reduced to a smaller number of factors, which can then be used for reporting or further analysis.

Outlines

00:00

📊 Introduction to Factor Analysis

Ruben from SPSS Tutorials introduces the concept of factor analysis, starting with a case study on unemployment agency performance in the Netherlands. The agency uses 16 survey questions to measure client satisfaction, scored on a 7-point scale. Ruben points out the limitations of simply computing mean scores for each statement and introduces the idea that some statements might measure the same underlying client sentiments, which is the fundamental concept behind factor analysis. He explains that factor analysis aims to reduce the number of observed variables into a smaller set of underlying factors, using a diagram to illustrate how nine statements might measure just three underlying factors.

05:04

🔍 Exploratory Factor Analysis (EFA) in SPSS

The tutorial delves into Exploratory Factor Analysis (EFA), which is used to identify the underlying structure of a set of observed variables. Ruben explains that EFA is one of two main types of factor analysis, the other being Confirmatory Factor Analysis (CFA). He clarifies that EFA is available in SPSS and is used to determine the likely factor model for a given dataset. The video outlines the three basic questions EFA aims to answer: the number of underlying factors, which items measure which factors, and what real-world characteristics the factors represent. Ruben also discusses how EFA starts by computing the pairwise correlations among all variables and uses scatter plots to visualize these correlations.

10:04

📈 Analyzing Correlations and Identifying Factors

Ruben demonstrates how to analyze correlations to identify underlying factors. He uses a hypothetical model where the first three items measure the same underlying factor, 'information clarity,' and shows that these items should have strong positive correlations. The video then compares this model to the actual correlations found in the data, using conditional formatting in Excel to visualize the correlation matrix. Ruben explains how the pattern of correlations can suggest the number of underlying factors and which items measure which factors. He also touches on the challenge of interpreting what the factors represent in the real world, using the example of items related to reliability and information clarity.

15:07

📋 Practical Considerations for Factor Analysis

The final paragraph addresses the practical aspects of conducting factor analysis. Ruben acknowledges the complexity of real-world datasets, which often require more sophisticated techniques due to large and intricate correlation matrices. He reiterates the use of conditional formatting to identify patterns in correlations and mentions the use of eigenvalues to determine the number of real factors. Ruben also discusses the interpretation of factor loadings, which indicate how strongly items measure certain factors, and the practice of adding actual factor scores to the dataset for reporting. The video concludes with a reminder to perform thorough data screening before running the analysis and encourages viewers to like and subscribe for more tutorials.

Keywords

💡Factor Analysis

Factor analysis is a statistical method used to describe variability among observed, correlated variables in terms of a potentially lower number of unobserved variables called factors. In the video, factor analysis is introduced as a way to understand how a set of variables might be underpinned by a smaller set of underlying factors. The example of the Dutch unemployment agency's client satisfaction survey is used to illustrate how factor analysis can help identify the underlying factors that contribute to client satisfaction.

💡Likert Scale

The Likert scale is a psychometric scale commonly used in research that asks respondents to report their level of agreement or disagreement with a statement on a scale. In the video, the unemployment agency's survey uses a 7-point Likert scale to measure client satisfaction, allowing for a nuanced understanding of responses that goes beyond simple yes/no answers.

💡Exploratory Factor Analysis (EFA)

Exploratory Factor Analysis (EFA) is a method of factor analysis that aims to identify the underlying structure of a relatively large set of variables. In the video, EFA is discussed as a way to explore the dimensionality of the client satisfaction survey data, with the goal of reducing the number of variables while retaining the maximum amount of information.

💡Confirmatory Factor Analysis (CFA)

Confirmatory Factor Analysis (CFA) is a statistical technique used to test the hypothesis that observed variables are related to one or more latent factors. Unlike EFA, which is exploratory, CFA is confirmatory and is used to test the validity of a factor structure. The video mentions CFA as a different type of factor analysis that is not available in SPSS, the statistical package used in the tutorial.

💡Eigenvalues

In the context of factor analysis, eigenvalues are numerical values that represent the amount of variance explained by each factor. The video script mentions that eigenvalues greater than one are considered to indicate significant factors, which is a common rule of thumb for determining the number of factors to retain in EFA.

💡Factor Loadings

Factor loadings are the correlations between the observed variables and the factors. They indicate the strength of the relationship between each variable and the underlying factors. In the video, factor loadings are discussed as a way to interpret which items measure which factors and to what extent.

💡Dimension Reduction

Dimension reduction is the process of reducing the number of random variables under consideration, by obtaining a set of principal variables. In the video, factor analysis is described as a method of dimension reduction, where the original 16 satisfaction statements are reduced to a smaller number of underlying factors.

💡Correlation Matrix

A correlation matrix is a table that shows the correlation coefficients between pairs of variables. In the video, the correlation matrix is used to identify patterns of correlations among the survey items, which can suggest the presence of underlying factors.

💡Ordinal Variables

Ordinal variables are data that have a natural order but the differences between the values are not necessarily equal. In the video, it is noted that the survey responses are ordinal variables, and while computing means on ordinal data is not strictly correct, it is a common practice in real-life research.

💡Client Satisfaction

Client satisfaction refers to the degree of contentment a client has with the services or products they have received. In the video, the main theme revolves around measuring client satisfaction through a survey and using factor analysis to understand the different aspects that contribute to it.

💡Conditional Formatting

Conditional formatting is a feature in spreadsheets like Excel that allows cells to be formatted based on their values. In the video, conditional formatting is used to visualize the correlation matrix by coloring cells based on the strength of the correlation, which helps in identifying patterns.

Highlights

Introduction to the basic idea behind factor analysis.

Case study on Dutch unemployment agency's client satisfaction survey.

Explanation of the unemployment agency's dual role in providing income and employment assistance.

Description of the 16 survey questions used to measure client satisfaction.

Discussion on the use of mean scores as a performance evaluation method.

Critique of using mean scores for ordinal variables in research.

Introduction to the concept of factor analysis for dimension reduction.

Illustration of how nine statements are believed to measure three underlying factors.

Differentiation between exploratory factor analysis (EFA) and confirmatory factor analysis (CFA).

Explanation of how EFA works to identify underlying factors from observed variables.

Discussion on the three basic questions EFA tries to answer.

Use of Pearson correlation to identify patterns in data for factor analysis.

Application of conditional formatting in Excel to visualize correlation patterns.

Interpretation of the correlation matrix to determine the number of underlying factors.

Introduction to eigenvalues as a method for determining the number of real factors.

Importance of factor loadings in interpreting the strength of relations between items and factors.

Practical application of factor analysis in reducing data dimensions for reporting.

Advice on conducting thorough data screening before running the actual factor analysis.

Transcripts

[Muziek]

Good morning everyone This Is SPSS

Tutorials My name is Ruben and in

Today's video I'll cover The Basic Idea

Behind Factor analysis and introduce Our

First case

Study for Our case Study we're going to

work on Doll

surs There No Need

download going to start off with some

very Basic Theory Now One Thing i' like

to show is that the Bulk of the data are

a bunch of statements so Let Me Give You

Just A Little Bit of background on

this so in the Netherlands we have an

unemployment Agency It's actually part

of the government and they have two Main

tasks so first off They need to provide

unemployed people with some monthly

income is know as the doll and second

They need to help unemployed people in

Finding employment

Again Now Our unemployment Agency Wants

To answer an important Question Which is

how well Are we performing and in order

to answer this They First formul 16

Survey questions on client Satisfaction

and Next up They approach A Simple

random sample of at

stats and all ofs and anwers data

file so Here we see the 16 statements

and each of them is scored on a 7 Point

lik scale with a no answer

Option Now Here we see some of the

Actual answers and one Option for

evaluating the Performance is to simply

compute the mean score over the 388

scores on v1 compute the mean score over

388 scores on V2 and So on Now as A

Minor Note some May Argue that These are

Strictly ordinal variables so computing

means SC is not really allowed but then

Again This is super Common in real life

Research and also it's the only

realistic Way To Go so in short for each

of these 16 statements

If Mean is Close

To and Mean is close to One That's

probably not such a good Mean score So

This Is One Option but it really Isn't A

Very Good Option so Let Me Give You A

few Reasons Why It

Isn't So first off There's a

Practical is separate me

scor client

Satisfaction so some of these statements

are going to be judged favora and some

other statements Maybe unf but how are

we going to summarize this into One

single Idea of client

Satisfaction Now second There's a More

philosophical Question whether client

Satisfaction really consists of 16 truly

separate

aspects and now finally reverse of is

the ide that Maybe some stats measure

the same underlying client sentiments

and this is The Very Basic Idea Behind

Factor

analysis Now this Basic Idea is nicely

illustrated by this diagram so first off

on the right Side we see nine Out Of The

16 statements and These are Simply The

variables in our data nou The Very Basic

Idea is that These nine statements

really Only measure three underlying

factors so for example the first three

statements are believed To All measure a

More general Sentiment of information

Clarity So This imp that If information

Clarity is improved then That's Going To

positively affect All three of these

statements and That's Why we see These

Arrow

indicate streng of Ars technically are

Simply B coefficients that we regr

analysis in short this model suggest

that n observed variables

pris isas we Factor analysis under

Dimension reduction

SPSS The Basic Idea is that we start off

with nine Dimensions The observed

statements in our data and we reduce

Them into three Dimensions die

underlying

factors however this Factor analysis

over here is Only One Of Two Main types

of Factor analysis and sta

There confirmat Factor analysis cfa for

Short and this is not available in

SPSS other

packages cfa basically answers to What

extent does a given factor modit

dat ex

a couple of Things for instance you want

to replace These Question Marks by

numbers that indicate the Strength Of

These

relations but also you probably want to

test whether this entire model actually

fits The data

colm May

typ of fact analysis is exploratory

Factor analysis e for Short and in SPSS

This Is Found under Factor EA Simply

answers What is a likely Factor model

some given data

set typically want to use EA If Only

dat No CLS

min technical is Origin Factor analysis

Only applied to cfa and not

EA and some statistici claim that EA is

not truly Factor

analysis They do have a Bit Of A Point

Here but Personally I think it's Way

More constructive to see EA and cfa As

M types of Factor

analysis so Anyway in this video we're

going to use exploratory Factor analysis

for Our case Study so Let's Now Zoom in

Just A Little Bit on what it does And

how it

works so first off efa Tries to answer

three Basic questions and the first of

these is how many underlying factors are

measured by k variables

Which are called items in Factor

analysis so Again for Our case Study The

items are These 16 Job Satisfaction

statements Right Question number two is

Which of these items measure Which

factors so for example We Could conclude

that Maybe the first Five items measure

Factor number one Maybe The Next Three

items measure Factor number

is a challenging one Which real world

characteristics do factors

represent for example if the first Five

items measure Factor Number One and

items are all about Communication

number May

Cent In This case we probably conclude

that Factor number represents some sort

of Sentiment on client friendliness and

So

on at this Point hopefully you wonder

How on earth Are we going to answer

difficult

starts by computing the pi correlations

Among All variables so as a very Quick

refresher a pieron correlation is a

number between min One and plus one that

indicates to what extent to quantitative

variables are linear

related now If that doesn't ring Any

Bell then let's take a Quick Look at

These scatter plots so for example over

here we see

correl of

0.7 Clear pattern is that if the scores

on variable a increase from left to

Right then The scores on variable B also

tend to increase and the extent to Which

this happens is the Pier correlation

Which is

0.7 this scatter plot over

here this May The

let Take Quick example model and Let's

for Now assume that This is correct so

One Thing we see is that the first three

items All measure the same underlying

Factor Which is information Clarity So

This imp that If information Clarity is

to increased the scores All

decre Clarity so in short the first

three items measuring the same

underlying Factor suggest That We should

find strong Positive correlations Among

These items this makes perfect Sense

right now the second three items measure

a Different underlying Factor Which is

Client friendliness

item items so in short this vor model

proposes a pattern of correlations Which

is that all correlations should be

Fairly Close To zer except for the

correlations Between these three items

The correlations These items Fin

correl items so Let's now take a look at

the Actual correlations and see if we

find this pattern or Maybe a Different

One So Here's the Actual

correlations but it's really hard to see

Any pattern whatsoever Because We Simply

have 81 numbers n if you work in Excel

There's a really nice Little trick Which

is conditional formatting color scales

And then we click the first Option Now

this applies background Colors That Run

from Dark Green for the highest number

in all of these cells to Dark Red for

the lowest number in all of these cells

Right And what we see Now is that we

have a sort of green Area over here

because all these correlations Among the

first three items are Rather high

however the first three items Don't

correlate very strongly with

let now revisit the Basic questions we

trying to answer with Factor

analysis so Again Question number one

how many underlying factors are measured

by k

items well like We Just this Green Area

indicates a group of variables that High

su that Meure underlying

Factor Here we One Green Area Here we

see a second and here we see a Third

Green Area So This pattern of

correlations suggest That We May be

measuring three underlying

factors Question number two is Which of

items Meure factors

under is

measured

items second underlying factor is

measured by the second three items and

finally The Third underlying factor is

measured By The Final three

items The Third and Final Question Is

Rather challenging It is Which real

world characteristics do factors

represent how are we to

pry Basic factors are Artificial

variables that represent Whatever

underlying items have in Common so let

this s in for Just Half A second and

with this in mind let's take Another

Final Look at Our

correlations so Again Factor number one

represents Whatever The items in

Common pry is

taken agreements Me Are followed through

and my contact person Always does What

he or She Promises So What do These

three items have in

Common say They All Con some element of

reliability so That's What the first

Factor probably

represents

simil items

pered Factor number and finally These

last three items All contain something

about information Clarity so That's

probably going to be Factor number

three Now Our correlation approach

nicely illustrat The Basic ide Behind

Factor

analysis real world datasets mostly

require powerful techniques A Reason

This is that we usually have to deal

with correlation Matrix that May be huge

and Fairly

complex so These are the Actual

correlations Among the 16 statements of

our First case Study obviously at this

Point we don't see Any pattern

whatsoever

but Remember my nice Little trick

conditional formatting Color scale

Green areas

Any This is the first Point Our

correlation approach is to fill with

real

life second and related Problem is If We

don't see Any very Clear Green Are

some No No I think These are Five so we

really need something More objective to

determine the Number of

factors so in a real life analysis we

compute Quality scores for Our factors

That are known as eigen

values to in later lecture Common of is

that factors with eigen value over One

are believed to be Real

factors This table suggests that There f

factors underlying Our 16 client

Satisfaction

statements a Third Point is that we want

to interpret Our factors by Looking Into

the correlations known as Factor loading

factors

items Factor loading and The Four colums

are Our Four factors dened as Component

Which is more or Less the same Thing as

a Factor but in Any case What's

important is that some items measure

some factors a bit more strongly than

other items and this is Something we

want to take into account When

interpreting The factors but Again wel

cover this in a Next

and Final is that Real analysis we Often

add The Actual Factor scores as new

variables to Our

data so Here we see the four factors for

Our First case Study and we could choose

The report On These instead of the 16

original statements and Again This Is

Why we find Factor analysis under

Dimension reduction in

SPSS Right so that will conclude this

video on some Basic ideas underlying

exploratory Factor

analysis Now before you Run The Actual

analysis You should First Do a thoro

data screening so That's What I'll cover

in the next video Anyway I Hope You

Found this video helpful please Give It

A thumbs up and subscribe to Our Channel

and if Any

[Muziek]

than watching see next time