Here’s a detailed, structured lesson on Factor Analysis in psychology—

including its types, how it’s performed, and how researchers interpret
the results.

🎯 What Is Factor Analysis?

Factor Analysis is a multivariate statistical technique used to identify
latent variables—underlying constructs such as intelligence, personality
traits, or attitudes—that explain the patterns of correlation among
observed variables (e.g. questionnaire items) (Verywell Mind).
Its main goals:
 Reduce data dimensions by summarizing many variables into a
few factors.
 Discover underlying constructs that cannot be measured directly.

🧭 Main Types of Factor Analysis

1. Exploratory Factor Analysis (EFA)
 Purpose: Uncover underlying structures without preconceived
hypotheses about how variables group together (ScienceDirect,
SurveySparrow, iResearchNet Psychology, Wikipedia).
 Process:
1. Choose extraction method (commonly Principal Axis
Factoring (PAF) or Maximum Likelihood (ML)).
2. Decide how many factors to retain (via scree plot, Kaiser
eigenvalue rule, or parallel analysis).
 3. Rotate factors (orthogonal like Varimax or oblique like
Promax) to enhance interpretability.
 Use-case: Early-stage research or scale development (Wikipedia,
ScienceDirect, Datamation).
2. Confirmatory Factor Analysis (CFA)
 Purpose: Test a pre-specified hypothesis about how variables
relate to latent factors (SurveySparrow).
 Process:
1. Specify model based on theory (which items load on which
factor).
2. Fit model using structural equation modeling.
3. Evaluate model fit indices (e.g., CFI, RMSEA, χ²) to judge
adequacy.
4. Constrain loadings or inter-factor correlations if theory
suggests independence or specific structure.
 Use-case: Later stages of scale development or theory testing
(Mental Health and Psychometrics, Wikipedia).
3. Principal Component Analysis (PCA) (sometimes grouped with
factor analysis)
 Purpose: Reduce data with the goal of explaining maximum total
variance, not necessarily uncover latent constructs.
 Key distinction: PCA extracts components, not true latent
common factors; better suited for data compression rather than
theoretical measurement (Datamation).
📊 Factor Extraction Methods
 Principal Axis Factoring (PAF): Extracts factors based on common
variance, not total variance. Useful when the goal is to infer latent
constructs. Less assumption-heavy than ML (Wikipedia).
 Maximum Likelihood (ML): Assumes data are normally distributed
and provides goodness-of-fit tests, confidence intervals, and
statistical tests of loadings—ideal when assumptions hold
(Wikipedia).
 PCA (technically not factor analysis but extraction technique):
Maximizes total variance, commonly used for dimensionality
reduction (SurveySparrow).

🧩 How to Decide How Many Factors to Retain

 Kaiser–Guttman rule: Retain factors with eigenvalues > 1.
 Scree plot: Graph eigenvalues, look for “elbow”—but subjective.
 Parallel analysis: Compare actual eigenvalues to randomized data;
retain factors exceeding simulated eigenvalues. Widely considered
reliable and more objective (FasterCapital, Wikipedia).

🔄 Rotation: Orthogonal vs. Oblique

 Orthogonal rotation (e.g. Varimax): Assumes factors are
uncorrelated; simplifies interpretation.
 Oblique rotation (e.g. Promax): Allows factors to correlate, often
more realistic in psychological constructs (Datamation, Reddit).
Rotation helps achieve simple structure, where each variable has high
loadings on one factor and low on others—making interpretation
clearer (ScienceDirect).

✅ EFA vs. CFA: Summary Table

Feature EFA CFA
None (data-
Hypothesis Pre-specified (theory-driven)
driven)
Explore potential Confirm and validate a known
Goal
factor structure structure
Any variable can 지정된 factor-variable assignments
Factor relationships
load on any factor 만 허용
Scale
Use-case development, Model testing and validation
theory generation
Loadings & Loadings, factor correlations, fit
Statistical output
rotated matrices indices
Separate sample for CFA to avoid
Sample split overfitting (Reddit, Mental
Same sample
recommendation Health and Psychometrics,
ScienceDirect, Reddit, Reddit)
Best Practice: Use EFA on one dataset to generate structure, then CFA
on an independent sample to confirm it (Reddit, Mental Health and
Psychometrics).
🧪 Practical Considerations & Assumptions
 Sampling Adequacy: Evaluate using the Kaiser–Meyer–Olkin
(KMO) test—values ≥ 0.80 are good; < 0.6 suggests factor analysis
may be inappropriate (Wikipedia).
 Sample Size: A common guideline is at least 5–10 observations
per variable, but many recommend 200+ cases as minimum;
larger samples improve stability and reliability (Reddit, Reddit,
Reddit).
 Assumptions: Linear relationships among variables, absence of
extreme multicollinearity or outliers, and (for ML) normality
(Spotfire, Reddit).

🧠 Why It Matters
Factor analysis helps:
 Simplify complex sets of variables into meaningful constructs.
 Provide evidence of construct validity by examining convergent
and discriminant relationships.
 Create and refine psychological scales with underlying theoretical
justification (iResearchNet Psychology, Wikipedia).

🔍 Walkthrough: Typical Steps in Conducting Factor Analysis

1. Assess suitability via KMO and Bartlett’s test.
2. Use EFA (PAF or ML) to extract factors.
3. Determine number of factors (scree plot, parallel analysis).
 4. Rotate factors (orthogonal or oblique) for interpretability.
5. Interpret loadings and label factors.
6. Confirm with CFA in separate dataset, check fit indices.
7. Make refinements: drop low-loading/redundant items, re-specify
model.

🧾 Final Summary
 EFA is used to explore latent structures when you have no firm
hypotheses.
 CFA is used to verify and validate a hypothesized structure.
 Extraction methods like PAF and ML differ in assumptions and
output.
 Rotation helps clarify factor interpretation.
 Ensure adequacy via KMO and sufficient sample size.
 Best practice: Explore with EFA, confirm with CFA using separate
data.
Let me know if you'd like help performing actual extraction, rotation,
interpretation steps, or looking at real-world examples in software like
SPSS, R, or MPlus!

Easy
Here’s a simple, easy-to-understand guide on Factor Analysis and its
main types in psychology:

🧩 What Is Factor Analysis?

Factor analysis is a statistical method that finds hidden patterns in a lot
of data. Imagine you have many survey questions—the analysis finds
that some of them go together and point to an underlying idea, or
factor, which you can’t see directly (Verywell Mind).
For example, answers to many personality questions may group into
basic traits like "extraversion" or "agreeableness"—that’s factor analysis
at work ([Link], Wikipedia).

🧭 Why Use It in Psychology?

 Simplifies data: Shrinks many observed variables into a few core
factors.
 Reveals hidden constructs: Helps measure things like intelligence,
personality, or attitudes that you can’t observe directly
(iResearchNet Psychology).
 Supports validity: Helps check if your test actually measures the
intended construct by seeing which items group together
(iResearchNet Psychology).

🌱 Two Main Types: EFA vs CFA

1. Exploratory Factor Analysis (EFA)
  Purpose: Discover unknown patterns in your data.
 When to use: You don’t know how many factors exist or which
items belong to which factor ([Link], SurveySparrow).
 How it works:
o Let the data reveal how items cluster.
o Decide how many factors to keep (using methods like scree
plots or eigenvalue rules).
o Rotate factors (e.g., Varimax or Promax) to make them
clearer and easier to interpret (Mental Health and
Psychometrics, Wikipedia, Reddit).
2. Confirmatory Factor Analysis (CFA)
 Purpose: Test a specific hypothesis—you define which items
should load on which factors.
 When to use: You already have a theory or structure and want to
confirm it's supported by the data (Mental Health and
Psychometrics).
 How it works:
o You specify a model in advance (which items go to which
factors, whether factors correlate, etc.).
o Fit the model and evaluate how well it matches the data
using model-fit stats (e.g., CFI, RMSEA) (Wikipedia).

🔍 EFA vs. CFA: Simple Comparison

Feature EFA CFA
Discover patterns without Confirm a pre-defined factor
Goal
assumptions structure
You decide which item loads
Structure Let the data decide
on which factor
Early research or scale Theory testing or validating a
Use case
development scale
Items can relate to any You control loading structure
Flexibility
factor (some loadings fixed)
Model Descriptive; no strict fit Uses fit indices to test model
evaluation indices adequacy

🧠 Working Flow
1. Check Data Suitability
Ensure sample size is enough (typically 100–200+ participants)
and variables are appropriate for factor analysis (e.g., not too
correlated or random) (TestSiteForMe, Reddit, Reddit, Reddit).
2. Run EFA first (if needed)
Explore and see how items cluster. Choose how many factors
make sense.
3. Rotate factors
Use orthogonal (Varimax) if you assume independent factors, or
oblique (Promax) if factors may correlate (Reddit, Reddit).
 4. Interpret factor loadings
Loadings show how strongly each item ties to a factor (e.g., > 0.4
suggests a strong link) (Reddit).
5. Confirm with CFA
Use a new dataset (or a split-half sample) to test the EFA findings.
CFA estimates model fit and tests your proposed structure
(Reddit).

✅ In a Nutshell
 EFA: Explore your data without assumptions, let items form
clusters.
 CFA: Test a specific pattern you expect—does the data agree?
 Use EFA first when you're developing scales; then use CFA to
validate your model on new data.

Easy Metaphor from Reddit

Think of EFA like watching where students naturally sit in the cafeteria
(forming groups/clusters). CFA is like laying out floor markers for groups
ahead of time and seeing if students actually follow that plan. If they do
—your model fits well (Reddit, Reddit, Reddit).

Would you like walking through an example with real data, or maybe
help using software like SPSS or R? I’d be glad to guide you!