Introduction
Have you ever collected survey data with dozens of questions and wondered how to make sense of it all? Looking at 30 separate answers is overwhelming. What you really want to know is whether the questions fall into a few simple themes.
That is exactly what exploratory factor analysis (EFA) does. You don’t need advanced math to understand it. This guide explains it step by step, with everyday examples and no unnecessary jargon.
What Is Exploratory Factor Analysis?
Exploratory factor analysis is a statistical method that groups related questions (variables) together to reveal the hidden themes behind them. These hidden themes are called factors.
Think of a school report card. Say a student gets high marks in algebra, geometry, and calculus. Those three scores move together because they all reflect one hidden skill: math ability. You can’t measure math ability directly, but you can see its effect on the three tests.
EFA does this automatically. You give it many variables, and it tells you which ones belong together and how many hidden themes there are.
A Simple Example
Imagine a survey with six statements, rated from 1 to 5:
I enjoy my daily tasks.
I feel proud of my work.
My salary is fair.
I am happy with my benefits.
My manager supports me.
My coworkers help me.
EFA might find three factors:
Factor 1: Job enjoyment (statements 1 and 2)
Factor 2: Pay and rewards (statements 3 and 4)
Factor 3: Workplace support (statements 5 and 6)
Instead of six separate results, you now have three clear themes.
Why Is It Called “Exploratory”?
Because you don’t know the answer in advance. You are exploring the data to see what structure appears. If you already had a theory and wanted to test it, you would use confirmatory factor analysis (CFA) instead.
In short:
EFA asks: “What patterns are in my data?”
CFA asks: “Does my expected pattern fit my data?”
When Should You Use EFA?
EFA is useful when you want to:
Create a new questionnaire and see how the questions group
Shorten a long survey by finding questions that measure the same thing
Understand the main themes in your data
Check whether the items in a scale really belong together
It is very common in psychology, education, health, business, and marketing research.
What You Need Before You Start
You don’t need to memorize these, but check them before running EFA:
Enough participants. Aim for at least 100 to 200 people, and 300 or more is better.
Related questions. The variables should be at least somewhat correlated. If they have nothing to do with each other, there are no themes to find.
The right type of data. Rating-scale questions (such as 1 to 5) work well.
Two quick tests in your software:
KMO test: checks whether your data is suitable. A value above 0.60 is acceptable, and above 0.80 is good.
Bartlett’s test: should be significant (p < .05), which means your variables are related enough to analyze.
How to Do EFA: Step by Step
Step 1: Prepare your data
Clean your data and fix missing values. If some questions are negatively worded (for example, “I often feel stressed” in a happiness survey), reverse their scores so all questions point the same way.
Step 2: Run the KMO and Bartlett’s tests
These confirm your data is suitable. If they fail, EFA may not be a good fit.
Step 3: Decide how many factors to keep
This is the trickiest step. Here are three simple tools:
Eigenvalue rule: keep factors with an eigenvalue above 1. (An eigenvalue is just a number showing how much information a factor captures.) This is easy but sometimes keeps too many.
Scree plot: a graph that drops like a hillside. Keep the factors before the line flattens out, at the “elbow.”
Parallel analysis: compares your results to random data. Experts consider it the most reliable method, so use it if your software allows.
Step 4: Choose a rotation
Rotation simply adjusts the results so they are easier to read. You have two main choices:
Varimax: assumes the factors are unrelated
Oblimin: allows the factors to be related
In most real-life surveys, themes are connected (for example, people who enjoy their work often feel supported too), so oblimin is usually a safe starting choice.
Step 5: Read the factor loadings
A factor loading tells you how strongly a question belongs to a factor. Think of it as a score from 0 to 1:
Loading What it means
Below 0.30 Weak, likely not useful
0.40 to 0.49 Acceptable
0.50 to 0.69 Good
0.70 or higher Excellent
A good question loads strongly on one factor and weakly on the others. If a question loads on two factors at once (a cross-loading) or on none, consider removing it.
Step 6: Name your factors
Look at the questions grouped under each factor and give the group a clear name, like “Job enjoyment” or “Workplace support.”
Step 7: Check reliability
Finally, calculate Cronbach’s alpha for each factor. It tells you whether the questions in a group are consistent with one another. A value of 0.70 or higher is generally considered acceptable.
How to Run EFA in Common Software
In SPSS:
Click Analyze > Dimension Reduction > Factor.
Add your questions to the Variables box.
Under Descriptives, select KMO and Bartlett’s test.
Under Extraction, choose Principal axis factoring and tick Scree plot.
Under Rotation, choose Direct Oblimin.
Under Options, tick Sort by size and hide small loadings below 0.30.
Click OK.
In R:
r
library(psych)
KMO(my_data)
fa.parallel(my_data, fa = "fa")
result <- fa(my_data, nfactors = 3, rotate = "oblimin", fm = "pa")
print(result, cut = 0.3)
Replace my_data with your data set and 3 with the number of factors you decided to keep.
Common Beginner Mistakes
Using too few participants. Small samples give unreliable results.
Trusting only the eigenvalue rule. Use the scree plot and parallel analysis too.
Mixing up EFA and PCA. They are similar but not the same. EFA looks for hidden themes, while PCA simply shrinks the data.
Removing many questions at once. Remove one at a time and re-run the analysis.
Forgetting to check reliability. Always test each factor with Cronbach’s alpha.
Not naming factors thoughtfully. A factor should make sense to a real reader.
Quick Glossary
Factor: a hidden theme behind a group of questions
Factor loading: how strongly a question connects to a factor
Eigenvalue: how much information a factor captures
Rotation: a step that makes results easier to read
Cross-loading: a question that loads on more than one factor
Communality: how much of a question’s information the factors explain
Frequently Asked Questions
Is EFA hard to learn?
The idea is simple, and software does the calculations. The main skill is making good decisions, such as how many factors to keep.
How many participants do I need?
At least 100 to 200, but 300 or more is safer.
What is the difference between EFA and PCA?
EFA finds hidden themes behind your questions. PCA just reduces many variables into fewer summary scores.
Which rotation should I choose?
Start with oblimin if you think your factors are related, which is common. Use varimax only if you want them to be completely unrelated.
Can I use EFA for any survey?
It works best for surveys with rating-scale questions and a reasonably large sample.
Conclusion
Exploratory factor analysis helps you turn a long list of questions into a few clear themes. Check that your data is suitable, decide how many factors to keep, choose a rotation, read the loadings, name your factors, and test reliability. Once you have followed these steps a couple of times, EFA stops feeling intimidating and becomes one of the most useful tools in your research toolkit.
Top comments (0)