Factor Analysis Service: Exploratory (EFA) and Confirmatory (CFA)
Find the structure behind your questionnaire items and show that your scales measure what they should. We run exploratory and confirmatory factor analysis and explain each decision.
Exploratory factor analysis (EFA)
EFA is used when you are not certain how your items group together, for example when developing or adapting a scale. We check that the data suit factor analysis (KMO and Bartlett’s test), choose the number of factors using parallel analysis, the scree plot and eigenvalues, and select an extraction method (principal axis factoring or maximum likelihood) and rotation (varimax for uncorrelated factors, promax or oblimin when factors are related). We then review loadings and cross-loadings, drop weak items with a clear rationale and name the factors.
Confirmatory factor analysis (CFA)
CFA tests whether a structure you expect from theory, or found in an earlier study, fits your data. We specify the measurement model, estimate it in AMOS, R (lavaan) or STATA, and report model fit with CFI, TLI, RMSEA and SRMR against accepted cut-offs. We examine standardised loadings, composite reliability and average variance extracted (AVE), and use modification indices sparingly and only when they make theoretical sense.
Construct validity evidence
Factor analysis is one of the main sources of evidence that a scale is valid. We show convergent validity (items load strongly on their own factor, AVE above 0.50) and discriminant validity (constructs are distinct, using Fornell-Larcker or HTMT). Pair this with a reliability check such as Cronbach’s alpha for each factor; our reliability and validity testing service covers both together.
Sample size and planning
Factor analysis needs enough respondents: a common minimum is 100 to 200 cases, or 5 to 10 respondents per item, and more when communalities are low. We advise on this before data collection. When your factors feed into a larger model of relationships, see structural equation modelling.
How the work runs
- You send the item-level data and the scale documentation.
- We check suitability, run EFA or CFA and compare competing structures.
- We document each decision on items, factors and fit.
- You receive loading tables, fit statistics, path diagrams and a methods paragraph.
Frequently asked questions
What is the difference between EFA and CFA?
EFA discovers the factor structure from the data. CFA tests a structure you specify in advance against the data.
How many factors should I keep?
We use several criteria together, mainly parallel analysis and the scree plot, plus whether the factors are interpretable, rather than relying on the eigenvalue-greater-than-one rule alone.
What are good fit values for CFA?
Common guides are CFI and TLI of about 0.90 to 0.95 or higher, RMSEA below about 0.08 and SRMR below about 0.08, interpreted alongside the design and sample.
Can you help me adapt a scale to a new context?
Yes. We can run EFA on the adapted version, then confirm the structure with CFA on a fresh sample where possible.
Which software do you use?
SPSS for EFA, and AMOS, R (lavaan) or STATA for CFA, depending on what you need.
Related statistical analysis services
Reliability & Validity Testing
Alpha, omega, test-retest and validity evidence for your scales.
Structural Equation Modelling (SEM & PLS-SEM)
CB-SEM and PLS-SEM for complex theories with several constructs.
Regression Analysis
Linear, logistic, moderation and mediation models with diagnostics.
Statistical Analysis Services
See every statistical method we support.
Also related: Likert scale analysis · Questionnaire design and sampling