KMO and Bartlett’s Test in SPSS: How to Run and Interpret It

Before you run a factor analysis in SPSS, there is one short check that can save you from a misleading result. The Kaiser-Meyer-Olkin (KMO) measure and Bartlett’s test of sphericity tell you whether your variables share enough common variance for factor analysis to make sense. Run them first, read them correctly, and you will know whether to proceed, tidy your variables, or rethink the analysis altogether.

This guide walks you through both tests in SPSS, click by click, with syntax you can paste, plain-language interpretation, and a template for reporting the results in APA style.

What Are the KMO and Bartlett’s Tests?

The KMO measure and Bartlett’s test are almost always reported together, because they answer two sides of the same question: are these variables suitable for factor analysis?

The KMO measure of sampling adequacy compares the size of the correlations between your variables with the size of the partial correlations. In plain terms, if your variables correlate with each other for reasons they share, rather than for reasons unique to each pair, factor analysis has something real to work with. KMO values range from 0 to 1, and higher is better.

Bartlett’s test of sphericity tests a different, simpler idea. It asks whether your correlation matrix is significantly different from an identity matrix — a matrix in which every variable correlates perfectly with itself and not at all with anything else. If Bartlett’s test is significant, your variables are related enough that factor analysis is worth attempting. If it is not significant, there is little shared structure for factors to capture.

Neither test proves that your factors will be meaningful. They are a suitability check, not a stamp of approval on your theory. A good KMO and a significant Bartlett’s test mean factor analysis is reasonable to try; a poor result means you should pause and investigate before going further.

When Should You Run Them?

Run KMO and Bartlett’s tests whenever you plan an exploratory factor analysis (EFA), and run them on the same set of items you intend to factor analyse. They are part of the factor analysis procedure in SPSS, not a separate analysis, so they appear in the factor analysis output alongside the communalities and the rotated solution.

They are also useful when you are checking a questionnaire before scoring it. If you are about to build scale scores from a set of items, a quick KMO and Bartlett’s check tells you whether those items hang together in the way a scale assumes. If they do not, you may need to drop weak items, check for coding errors, or reconsider whether the items belong on one scale at all.

Do not run these tests as a substitute for thinking about your items. A high KMO on a set of items that measure unrelated things is unlikely, but a borderline KMO on a thoughtful, well-designed scale still deserves a closer look at individual items rather than an automatic rejection.

What You Need Before You Start

You need a data file in which each variable you want to factor analyse is a separate column, with one row per respondent. The variables should be measured at interval or ratio level, or treated as such, as is common with Likert-type questionnaire items in practice.

Sample size matters, though there is no single magic number. Factor analysis generally works better with larger samples, and very small samples can produce unstable results even when KMO looks acceptable. As a rough working guide, aim for a comfortably larger sample than the number of items, and be cautious about interpreting factors from a very small group. Check for obvious data problems first: miscoded values, reversed items that have not been reverse-scored, and variables with almost no variation will all distort the picture.

You do not need to standardise your variables first. SPSS handles the correlation matrix behind the scenes, so you can work with the raw item scores.

How to Run the KMO and Bartlett’s Test in SPSS: Step by Step

Follow these steps exactly, and you will get both tests in one run.

1. Open your data file in SPSS.
2. Go to Analyze > Dimension Reduction > Factor.
3. In the Factor Analysis dialog, move all the variables you want to factor analyse from the left-hand list into the Variables box on the right.
4. Click Descriptives. In the Statistics section, tick KMO and Bartlett’s test of sphericity. You can also tick Anti-image here if you want the individual measures of sampling adequacy for each variable — these are genuinely useful when the overall KMO is low, so ticking this box is worth it.
5. Click Continue to return to the main Factor Analysis dialog.
6. If you want to see the full factor solution as well, leave the default extraction settings for now. If you only want the suitability tests, that is fine too — the KMO and Bartlett’s results appear regardless of the extraction options.
7. Click OK.

SPSS will produce the factor analysis output. The KMO and Bartlett’s test table is near the top, just after the descriptive statistics, so you do not need to scroll far to find it.

SPSS Syntax for the KMO and Bartlett’s Test

If you prefer syntax, or you want a record of exactly what you ran, paste the block below into a Syntax window, replace the variable names with your own, and run it. The KMO and PRINT subcommands are what request the two tests.

FACTOR
/VARIABLES item1 item2 item3 item4 item5 item6
/MISSING LISTWISE
/ANALYSIS item1 item2 item3 item4 item5 item6
/PRINT INITIAL KMO EXTRACTION ROTATION
/CRITERIA MINEIGEN(1) ITERATE(25)
/EXTRACTION PC
/ROTATION VARIMAX
/METHOD=CORRELATION.

The /PRINT KMO part is the key instruction. Without it, SPSS runs the factor analysis but does not print the KMO and Bartlett’s table. Keeping the syntax also makes your analysis reproducible: you, a supervisor, or a reviewer can rerun exactly the same check on the same variables later.

How to Interpret the KMO Value

SPSS reports a single overall KMO value in the KMO and Bartlett’s Test table. Use these widely used guidelines to judge it:

– 0.90 and above: marvellous — excellent sampling adequacy.
– 0.80 to 0.89: meritorious — good.
– 0.70 to 0.79: middling — acceptable for most purposes.
– 0.60 to 0.69: mediocre — proceed with caution.
– 0.50 to 0.59: miserable — weak, and you should investigate before proceeding.
– Below 0.50: unacceptable — factor analysis is not appropriate for these variables as they stand.

Treat these bands as guidance, not as pass-or-fail cut-offs written in stone. A KMO of 0.68 is not automatically a disaster, and a KMO of 0.71 is not automatically fine. Look at the value, look at the individual measures, and look at your items.

The individual measures of sampling adequacy, found on the diagonal of the anti-image correlation matrix, show the same idea variable by variable. If the overall KMO is low, scan this diagonal: variables with individual values below 0.50 are the most likely culprits. Removing or fixing one weak item sometimes lifts the overall KMO noticeably, but remove items for substantive reasons — a poorly worded or miscoded item — not simply to chase a higher number.

Also check the off-diagonal values in the anti-image matrix. Large partial correlations there suggest pairs of variables that share little beyond what the factors explain, which is another sign that the set may need tidying.

How to Interpret Bartlett’s Test of Sphericity

Bartlett’s test gives you three numbers: an approximate chi-square value, degrees of freedom, and a significance (p) value. The one that matters for your decision is the significance value.

If Bartlett’s test is significant — conventionally, p < .05 — you can reject the idea that your correlation matrix is an identity matrix. In practical terms, your variables are sufficiently correlated for factor analysis to be worth attempting.If Bartlett's test is not significant (p >= .05), factor analysis is unlikely to be useful for this set of variables, because there is no shared correlational structure for factors to summarise. Check for data errors first, then reconsider whether these items belong together at all.

One caution: Bartlett’s test is sensitive to sample size. With a very large sample, even trivial correlations can produce a significant result, so a significant Bartlett’s test on its own does not guarantee a good factor solution. That is exactly why it is reported alongside KMO: Bartlett’s test tells you the correlations are not all zero, and KMO tells you whether they are substantial enough to be useful.

What to Do If Your Results Are Poor

A low KMO or a non-significant Bartlett’s test is information, not a dead end. Work through these checks in order.

First, check your data. Look for coding errors, items that were not reverse-scored, and variables with near-zero variance. A single miscoded item can drag the whole set down.

Second, look at the individual measures of sampling adequacy and the correlation matrix. Drop or revise the weakest items, one at a time, and rerun the check after each change. Removing several items at once makes it hard to tell which change helped.

Third, check your sample. If your sample is very small relative to the number of items, the honest answer may be to collect more data rather than to force a factor solution out of too little information.

Fourth, reconsider the item set. If the items genuinely measure different things, no statistical fix will turn them into a coherent factor. Splitting the set, or revising the questionnaire, may be the right move.

Do not report only the rerun that finally looks good while hiding earlier attempts. Report the final item set and the KMO and Bartlett’s results for that set, and be ready to explain any items you removed.

Reporting in APA Style (7th ed.)

Report the KMO and Bartlett’s results in your Method or Results section, before you describe the factor solution. You do not need a table for these two tests unless your institution requires one; a sentence or two in the text is standard.

A template you can adapt is:

“The Kaiser-Meyer-Olkin measure verified the sampling adequacy for the analysis, KMO = .82, and Bartlett’s test of sphericity indicated that correlations between items were sufficiently large for factor analysis, χ²(45) = 612.40, p < .001."Replace the KMO value, the degrees of freedom, and the chi-square value with your own output, and report the exact p value if SPSS gives one above .001. Note the degrees of freedom in parentheses after the chi-square symbol, and keep the same number of decimal places throughout your results section for consistency.If you removed items after an initial check, say so briefly, for example: "Two items with individual measures of sampling adequacy below .50 were removed before the final analysis." Readers and reviewers value that transparency far more than a silently perfect KMO.

Key Takeaways

  • Run the KMO and Bartlett’s tests before interpreting any factor analysis, on the exact items you plan to analyse.
  • KMO tells you whether the correlations are substantial enough; Bartlett’s test tells you whether they differ from zero. You need both.
  • Aim for KMO of .70 or above for comfortable interpretation, treat .60 to .69 with caution, and treat below .50 as a stop sign.
  • A significant Bartlett’s test (p < .05) supports proceeding; a non-significant result suggests factor analysis will not help this item set.
  • When results are poor, check coding, inspect individual sampling adequacy values, and remove weak items one at a time for substantive reasons.

Frequently Asked Questions

Do I need a significant Bartlett’s test AND a good KMO?

Yes, ideally. A significant Bartlett’s test without an acceptable KMO can happen with a large sample, where even small correlations reach significance. A good KMO with a non-significant Bartlett’s test is rare and usually points to a small sample or an unusual correlation pattern. Report both, and be cautious if either one is poor.

What is a good KMO value in SPSS?

Values of .80 and above are generally considered good, .70 to .79 acceptable, and .60 to .69 mediocre. Below .50, factor analysis is not appropriate for the variables as they stand. Use these bands as guidance alongside the individual item values, not as a rigid pass mark.

Can I run factor analysis if Bartlett’s test is not significant?

Technically you can run it, but the result is unlikely to be meaningful, because a non-significant Bartlett’s test suggests your variables share too little correlation structure for factors to summarise. Check your data and item set first, and consider whether factor analysis is the right analysis for these variables.

Where is the KMO test in SPSS?

Go to Analyze > Dimension Reduction > Factor, move your variables into the Variables box, click Descriptives, and tick “KMO and Bartlett’s test of sphericity”. The results appear in the KMO and Bartlett’s Test table at the top of the factor analysis output.

Does the KMO test require normally distributed data?

KMO itself is based on correlations and does not carry a strict normality assumption in the way some other tests do. Bartlett’s test, however, is sensitive to departures from multivariate normality, and factor analysis results in general are more trustworthy when your data are reasonably well behaved. Checking distributions before factor analysis is still good practice.

Need this done on your data? Get a free quote → /services/