How to run factor analysis in SPSS for a nursing questionnaire is easier when every decision follows the research question rather than the software defaults. Exploratory factor analysis (EFA) can show whether questionnaire items cluster into meaningful latent dimensions, but valid interpretation requires more than clicking Analyze. This guide covers data screening, KMO and Bartlett’s test, extraction, factor retention, scree plots, rotation, loadings and nursing-dissertation reporting.
What factor analysis does in a nursing questionnaire
Factor analysis examines patterns of correlations among observed questionnaire items and estimates whether a smaller number of underlying factors may explain their shared variation. For example, twelve items about nurses’ working conditions might reflect dimensions such as leadership support, workload pressure and team communication. Those dimensions are not measured directly; they are inferred from how item responses relate.
EFA is normally used when the dimensional structure is uncertain or when a scale is being explored in a new population or setting. Confirmatory factor analysis (CFA), usually conducted in software such as AMOS, R or Mplus, tests a prespecified measurement model. Do not describe an exploratory SPSS procedure as CFA.
Factor analysis is also different from reliability analysis. Cronbach’s alpha estimates internal consistency under particular assumptions; it does not discover the factor structure or establish validity by itself. Run reliability analyses separately after defining defensible subscales. This article therefore keeps Cronbach’s alpha outside the main analysis.
Before you run factor analysis in SPSS
Confirm that the questionnaire and research aim suit EFA
Use EFA for a set of items expected to reflect latent constructs. Do not automatically factor analyse demographic variables, unrelated clinical measurements or a collection of single-item outcomes. Items should have a conceptual reason to belong in the same candidate measurement model.
Decide whether the instrument is newly developed, adapted, translated or being tested in a different nursing population. State why its structure needs exploration. If a well-established scale has a firmly specified model, CFA may be more appropriate than rediscovering a structure with EFA.
Check sample size and response quality
No single participants-per-item rule guarantees an adequate sample. Stability depends on communalities, factor strength, number of items per factor, missingness and distribution. Justify the achieved sample using methodological literature and the characteristics of your dataset rather than stating that “five respondents per item is always enough”. Larger samples generally produce more stable correlation estimates.
Before analysis, check:
- one row represents one participant and one column represents one item;
- item labels and value labels are correct;
- reverse-worded items have been recoded correctly;
- out-of-range values and duplicate cases have been investigated;
- missing-data handling is stated and justified;
- items have sufficient variation and are not almost constant;
- the effective analysis sample is reported.
SPSS commonly uses listwise exclusion in factor analysis when that option is selected, so the analysis N may be smaller than the number of submitted questionnaires. Report the actual N used. For a broader preparation process, use the site’s guide to cleaning nursing research data in SPSS.
Inspect the correlation matrix
EFA depends on relationships among items. A matrix containing few meaningful correlations may not support a common-factor model. Conversely, correlations that are extremely high can indicate redundancy or multicollinearity. Request the correlation matrix and determinant in SPSS, inspect unexpected negative relationships and revisit reverse coding where necessary.
How to run factor analysis in SPSS for a nursing questionnaire: menu steps
- Open the cleaned SPSS dataset.
- Select Analyze > Dimension Reduction > Factor.
- Move only the questionnaire items intended for the same analysis into the Variables box.
- Select Descriptives. Request the initial solution, KMO and Bartlett’s test, anti-image matrix, correlation matrix and determinant where relevant.
- Select Extraction. Choose an extraction method justified by the aim, request the scree plot and set an initial factor-retention rule.
- Select Rotation. Choose an oblique or orthogonal method based on theoretical and empirical expectations.
- Select Options. Sort coefficients by size and suppress small coefficients only for readability; retain the unsuppressed output for checking.
- Choose the missing-value approach specified in the analysis plan.
- Select OK, then interpret the output as a connected sequence rather than isolated tables.
Save the SPSS syntax using Paste instead of relying only on the dialog boxes. Syntax creates a reproducible record of the variables and settings. Adapt the following illustration rather than copying item names blindly:
FACTOR
/VARIABLES q1 q2 q3 q4 q5 q6 q7 q8 q9 q10 q11 q12
/MISSING LISTWISE
/ANALYSIS q1 q2 q3 q4 q5 q6 q7 q8 q9 q10 q11 q12
/PRINT INITIAL KMO EXTRACTION ROTATION
/PLOT EIGEN
/CRITERIA ITERATE(100)
/EXTRACTION PAF
/ROTATION OBLIMIN
/METHOD=CORRELATION.Step 1: Interpret KMO and Bartlett’s test
The Kaiser–Meyer–Olkin (KMO) measure evaluates whether patterns of correlations are sufficiently compact for common factors to be useful. It ranges from 0 to 1, with higher values indicating better sampling adequacy. UCLA’s annotated SPSS guidance presents .60 as a suggested minimum, but a threshold should not replace judgement. Examine the overall KMO alongside item-level measures in the anti-image correlation matrix.
Bartlett’s test of sphericity tests the null hypothesis that the item correlation matrix is an identity matrix. A statistically significant result indicates that the correlations, taken together, differ from zero and therefore meet a minimum requirement for factor analysis. Do not write that Bartlett’s test “proves the questionnaire is valid”. With a large sample, small departures may become significant, so interpret it together with KMO, correlations and the factor solution.
| Output | Question answered | Cautious interpretation |
|---|---|---|
| Overall KMO | Is the shared correlation pattern broadly suitable? | Higher values support factorability; justify the threshold used. |
| Anti-image diagonal | Does each item show adequate sampling adequacy? | Low item values may identify problematic items, but removal also needs conceptual justification. |
| Bartlett’s test | Is the correlation matrix distinguishable from an identity matrix? | A significant result supports proceeding; it does not establish construct validity. |
| Determinant | Is extreme multicollinearity evident? | A value near zero may indicate computational or redundancy concerns. |
A dissertation sentence might read: “The 12 items were suitable for exploratory factor analysis because the overall KMO was .82 and Bartlett’s test was statistically significant, χ²(66) = [value], p < .001.” Replace placeholders with your actual output and add any relevant item-level concerns.
Step 2: Choose the extraction method
SPSS defaults can encourage students to confuse principal components analysis (PCA) with common factor analysis. PCA summarises total variance into components. EFA methods such as principal axis factoring (PAF) model shared variance to estimate latent factors. If the aim is to explore underlying constructs in a nursing questionnaire, PAF is often more conceptually aligned than PCA. Maximum likelihood factor analysis is another option when its distributional assumptions and inferential benefits are appropriate.
State the method and why it matches the aim. Avoid claiming that PAF is universally superior. A good justification identifies the construct-focused purpose, distributional considerations and planned interpretation.
Interpret communalities
Communality represents the proportion of an item’s variance explained by the retained common factors. Low extraction communalities suggest that an item is not well represented by the solution. Do not remove an item automatically because it falls below one universal cut-off. Consider the sample, loading pattern, content importance, number of items supporting the factor and the effect of removal on interpretability.
Step 3: Decide how many factors to retain
Factor retention is one of the most consequential decisions. Do not rely solely on SPSS’s eigenvalue-greater-than-one default. Use converging evidence:
- the questionnaire’s theory and intended constructs;
- the scree plot;
- eigenvalues and explained common variance;
- parallel analysis where available;
- the number and strength of items per factor;
- simple structure and clinical interpretability;
- comparison of plausible neighbouring solutions.
Read the scree plot correctly
The scree plot graphs eigenvalues against factor number. Look for the point where the steep decline levels into a flatter tail. Factors before that elbow are candidates for retention. When the elbow is ambiguous, report that uncertainty and compare alternative solutions rather than choosing the plot that supports a preferred result.
Parallel analysis compares observed eigenvalues with those produced from random data and is generally a stronger retention aid than the eigenvalue-greater-than-one rule alone. Standard SPSS dialogs may not provide it directly, so follow an approved procedure or extension and describe it transparently. Do not claim that the scree plot alone “confirmed” the final model.
Step 4: Select and justify factor rotation
Rotation makes the loading pattern easier to interpret. Orthogonal methods such as Varimax constrain factors to be uncorrelated. Oblique methods such as Direct Oblimin or Promax allow factors to correlate. Nursing constructs—such as workload, leadership support and team functioning—often have plausible relationships, making an oblique solution a reasonable starting point.
Use Varimax only when an uncorrelated-factor assumption is defensible, not because it produces a shorter table. If an oblique rotation shows negligible factor correlations, you may compare an orthogonal solution. Document the decision and avoid switching methods solely to obtain cleaner results.
| Rotation | Assumption | Main output for interpretation |
|---|---|---|
| Varimax | Factors are constrained to be uncorrelated. | Rotated Factor Matrix |
| Direct Oblimin | Factors may correlate. | Pattern Matrix, Structure Matrix and Factor Correlation Matrix |
| Promax | Factors may correlate; useful in many larger item sets. | Pattern Matrix, Structure Matrix and Factor Correlation Matrix |
Step 5: Interpret factor loadings and cross-loadings
A factor loading indicates the relationship between an item and a factor. Its sign shows direction and its magnitude shows strength. With oblique rotation, use the pattern matrix for the primary interpretation while also checking the structure and factor-correlation matrices.
Set a loading threshold in advance with reference to sample size and methodological guidance. Values such as .30 or .40 are often used as display or interpretation rules, but they are not universal validity boundaries. An item can load strongly yet be conceptually inconsistent with the other items.
A cross-loading item has meaningful loadings on more than one factor. Consider:
- the size of the primary loading;
- the difference between primary and secondary loadings;
- the wording and content of the item;
- whether reverse wording caused method effects;
- whether removal weakens content coverage;
- whether the factor remains supported by enough items.
Item deletion should be iterative and theoretically defensible. Remove one problematic item at a time where justified, rerun the complete analysis and record how KMO, communalities, factor retention and loadings changed. Do not delete items simply to maximise explained variance.
Step 6: Name the factors without overclaiming
Name each factor from the shared meaning of its strongest items, not from one item or from the label you hoped to find. Keep labels specific. “Perceived ward leadership support” is more defensible than “quality” when items concern manager availability, recognition and communication.
Check whether negatively worded items form a separate factor. Such clustering may reflect wording method rather than a genuine nursing construct. Discuss alternative explanations and avoid treating every statistical cluster as a new theoretical discovery.
Worked nursing questionnaire interpretation example
Assume a hypothetical 12-item questionnaire explores newly qualified nurses’ workplace support. KMO is .84 and Bartlett’s test is significant. PAF and the scree plot suggest three factors, while an oblique rotation shows modest correlations among them. Items about preceptor availability, feedback and approachability load on Factor 1; items concerning staffing pressure and missed breaks load on Factor 2; and items about speaking up and team respect load on Factor 3.
The factors might be labelled “preceptorship support”, “workload pressure” and “team psychological safety”. If one item about asking for assistance loads .44 on Factors 1 and 3, the researcher should examine its meaning and cross-loading difference instead of placing it automatically. The item may genuinely bridge preceptorship and speaking-up climates.
This example is illustrative, not a result to copy. Your dissertation must report the actual sample, instrument, output and decisions.
How to report factor analysis in a nursing dissertation
Report the analysis so a reader can understand and evaluate every major decision. Include:
- the reason EFA was appropriate;
- the questionnaire items and analysis sample;
- data screening and missing-data treatment;
- KMO, Bartlett’s test and relevant correlation diagnostics;
- extraction method and justification;
- factor-retention evidence, including scree plot and any parallel analysis;
- rotation method and justification;
- loading and cross-loading criteria;
- the final pattern or rotated matrix;
- factor names, item content and explained variance;
- items removed and the rationale for each decision;
- limitations and the need for validation in another sample.
Keep the main tables readable and place extensive SPSS output in an appendix if your university permits. Do not paste an entire output viewer into the results chapter without explanation. The site’s guide to discussing SPSS results in a nursing dissertation explains how to connect statistical findings to the research question.
Example results paragraph
“Exploratory factor analysis using principal axis factoring was conducted on 12 workplace-support items (n = [N]). Sampling adequacy was supported by KMO = [value], while Bartlett’s test was significant, χ²([df]) = [value], p < .001. Factor retention considered the scree plot, [parallel analysis/theory] and solution interpretability. A three-factor Direct Oblimin solution was retained because the constructs were expected to correlate. The factors were labelled [names] and accounted for [value]% of the common variance. Table X presents pattern coefficients above [threshold].”
Common SPSS factor-analysis mistakes
- Calling PCA an exploratory common-factor analysis without justification.
- Using eigenvalues above one as the only retention rule.
- Selecting Varimax automatically when constructs may correlate.
- Reporting only “KMO was acceptable” without its value.
- Claiming significant Bartlett’s test proves validity.
- Deleting cross-loading items without examining their content.
- Interpreting an oblique structure matrix as though it were the pattern matrix.
- Hiding the loading threshold or suppressed coefficients.
- Naming factors before considering all high-loading items.
- Reporting a solution without the effective sample or missing-data rule.
- Treating EFA in one sample as final validation.
Getting support with nursing questionnaire factor analysis
Factor analysis involves connected decisions, so support should explain and document the reasoning rather than produce unexplained SPSS tables. Nursing Dissertation Service can help nursing students check data preparation, choose defensible EFA settings, interpret KMO and Bartlett’s test, compare factor solutions and present results clearly. You remain responsible for the research decisions and must follow your university’s academic-integrity rules.
Use the contact page to describe your questionnaire, item count, sample, research aim and permitted support. Remove patient-identifiable, staff-identifiable or confidential organisational information before sharing data.
Frequently asked questions
What KMO value is acceptable for factor analysis?
Around .60 is often presented as a minimum, with higher values preferable. Interpret the overall KMO alongside item-level sampling adequacy, correlations, sample characteristics and the stability of the solution.
Should I use PCA or principal axis factoring?
Use PCA when the aim is component-based data reduction. Use a common-factor method such as PAF when the aim is to explore latent constructs represented by shared item variance. State the purpose and justify the choice.
Should nursing questionnaire factors be rotated with Varimax or Oblimin?
Oblique rotation is appropriate when factors may correlate, which is plausible for many nursing and healthcare constructs. Varimax assumes uncorrelated factors. Base the decision on theory and observed factor correlations.
How many factors should I retain?
Use converging evidence from theory, scree plot, parallel analysis where available, eigenvalues, item support and interpretability. Do not use the eigenvalue-greater-than-one rule alone.
Does factor analysis prove that my questionnaire is valid?
No. EFA provides evidence about internal structure in a particular sample. Construct validity requires a broader body of evidence, and the structure should be tested in an independent sample where possible.
Conclusion
Knowing how to run factor analysis in SPSS for a nursing questionnaire requires more than producing a rotated matrix. Begin with a defensible measurement aim and clean item data, assess factorability using KMO, Bartlett’s test and correlations, justify extraction and rotation, and retain factors using several sources of evidence. Interpret loadings with item meaning in mind, document every deletion and report the solution cautiously so readers can judge its relevance to nursing research.
References
IBM (2026) IBM SPSS Statistics documentation. Available at: https://www.ibm.com/docs/en/spss-statistics/ (Accessed: 28 August 2026).
UCLA Institute for Digital Research and Education (n.d.) Factor analysis: SPSS annotated output. Available at: https://stats.oarc.ucla.edu/spss/output/factor-analysis/ (Accessed: 28 August 2026).
Watkins, M.W. (2018) ‘Exploratory factor analysis: A guide to best practice’, Journal of Black Psychology, 44(3), pp. 219–246. doi: 10.1177/0095798418771807.
Williams, B., Onsman, A. and Brown, T. (2010) ‘Exploratory factor analysis: A five-step guide for novices’, Australasian Journal of Paramedicine, 8(3). doi: 10.33151/ajp.8.3.93.
