SPSS command: SPSS MVA

Missing value analysis (MVA) in your browser

Q: Little's MCAR test is significant. What now?

The data are probably not missing completely at random, so listwise deletion may bias results. Look at the separate-variance t tests to see which observed variables predict missingness, include them in the model, and use EM or Multiple Imputation, which assume MAR rather than MCAR.

Q: Does a non-significant MCAR test prove the data are MCAR?

No. It only means there is no evidence against MCAR in these variables; with few cases the test has little power, and it cannot detect missingness that depends on the missing values themselves.

Q: Can I use the EM means as my final results?

They are good point estimates under MAR, but they come without standard errors that reflect the missing information. For tests and confidence intervals, impute several datasets and pool them with Rubin's rules.

Q: Are user-missing codes such as 99 treated as missing?

Yes. Codes declared as user-missing in the variable metadata are counted as missing, and a footnote lists how many there were per variable.

Q: Is my data uploaded?

No. MVA runs in a web worker inside your browser tab; the file is never sent to Crosstabs servers.

Missing Value Analysis tells you how much data is missing, in which combinations, and whether the gaps look random.

Try with example data Run on your own file

Free · No account · Files stay on your device · Updated 18 October 2026

What it tells you

Crosstabs runs the SPSS MVA procedure on your device: univariate missing counts with extremes, tabulated patterns, separate-variance t tests, listwise, pairwise and EM estimates, and Little's MCAR test.

How many values each variable is missing, as a count and a percentage, with system-missing and declared user-missing codes both counted.
Which combinations of variables tend to be missing together, and how many cases each pattern holds.
Whether people who skipped one question differ on the others: a Welch t test compares the present and missing groups on every quantitative variable.
Whether the data are consistent with missing completely at random (MCAR), via Little's chi-square test computed from the EM estimates.
How means, covariances and correlations change between listwise deletion, pairwise deletion and EM (maximum likelihood under MAR).

When to use it

Use it when

Before any model on survey or panel data, to decide whether complete-case analysis is defensible or imputation is needed.
When a reviewer or client asks whether the missingness is random: report Little's MCAR test and the separate-variance t tests.
To get EM means and correlations as a quick MAR-based summary when a full multiple imputation is more than you need.

Use something else when

To produce a filled-in dataset for analysis: EM estimates are summary statistics, not imputed values. Use Multiple Imputation and pool the results.
To fill gaps in a time series by neighbouring values: use Replace Missing Values.
With a sampling (survey) weight when you need tests: only the descriptive tables are shown, so use Complex Samples for design-based inference.

How to run it in Crosstabs

01
Open your data
Open a .sav, CSV or Excel file in the workspace and switch to Statistics mode. Declared user-missing codes from a .sav file are respected.
02
Choose Missing Value Analysis
In the Missing values group, pick Missing Value Analysis (MVA).
03
Pick the variables
Tick the Quantitative variables (at least one) and, optionally, Categorical variables; categorical ones are counted in the missing statistics and patterns but are not part of EM.
04
Choose the tables
Keep Tabulated patterns and chart and Separate variance t tests on, and set Omit patterns with less than (% of cases) and Omit indicators missing less than (%) (defaults 1% and 5%). Tick EM estimates and Little's MCAR test with Maximum EM iterations, and Listwise estimates or Pairwise estimates if you want to compare them.
05
Declare the weight, then run
If a weight is active, say whether it is a frequency weight or a sampling weight (it is never guessed), then press Run. The analysis runs in a background worker on your device.

Run Missing Value Analysis in your browser

Worked example

Computed by the Crosstabs engine from the synthetic data below when this page was built — not typed in.

Try with example data

Data

60 synthetic customers with age, income (thousands), satisfaction (0–10) and region. Income is deliberately left blank more often for customers under 35, so the missingness depends on an observed variable; satisfaction and region are blanked at random. A fixed seed makes every build identical.

Univariate Statistics

Item	N	Mean	Std. Deviation	Missing Count	Missing Percent	No. of Extremes Low	No. of Extremes High
age	60	41.63	11.31	0	0.0%	0	0
income	44	56.65	11.54	16	26.7%	0	0
satisfaction	53	6.22	1.33	7	11.7%	1	0
region	56			4	6.7%

Extremes: cases outside Q1 − 1.5 × IQR and Q3 + 1.5 × IQR.

Separate Variance t Tests

Missing indicator	Variable	t	df	P(2-tail)	# Present	# Missing	Mean (Present)	Mean (Missing)
income	age	4.42	29.2	<.001	44	16	44.91	32.63
income	satisfaction	0.49	17.9	.629	40	13	6.27	6.05
satisfaction	age	3.28	13.1	.006	53	7	42.68	33.71
satisfaction	income	1.60	4.3	.180	40	4	57.31	49.98
region	age	0.63	3.3	.571	56	4	41.93	37.50
region	income	0.06	2.1	.958	41	3	56.69	56.07
region	satisfaction	0.89	3.6	.429	49	4	6.26	5.70

EM Means

Item	age	income	satisfaction
Mean	41.63	53.80	6.18
Std. Deviation	11.31	12.39	1.32

Little's MCAR test: Chi-Square = 16.299, DF = 5, Sig. = .006. EM converged after 13 iterations.

How to read it

23 of 60 cases (38.3%) have at least one missing value. The separate-variance t test shows that customers with missing income are younger (mean age 32.6 vs 44.9, p < .001), and Little's MCAR test gives χ²(5) = 16.30, p = .006. Read together, they warn that a complete-case analysis would over-represent older customers; the EM mean of income (53.80, against 56.65 for the observed values) uses age to adjust for that.

Output tables explained

Univariate Statistics: N, mean, standard deviation, missing count and percent, and the number of low and high extremes (outside Q1 − 1.5 × IQR and Q3 + 1.5 × IQR, or mean ± 2 SD for large files).
Separate Variance t Tests: For each indicator variable, cases are split into present and missing; each quantitative variable is compared with a Welch t test. Small p values point away from MCAR.
Tabulated Patterns: Each row is a missing-value pattern (X marks a missing variable) with its number of cases and a Complete if… count: how many cases would be complete if those variables were dropped. The Missing Value Patterns chart shows the same patterns.
EM Means: Means and standard deviations estimated by the EM algorithm; the footnote carries Little's MCAR test (chi-square, df, Sig.) and the convergence status. EM Covariances and EM Correlations follow.
Summary of Estimated Means and Standard Deviations: Listwise: Optional listwise and pairwise tables (means, covariances, correlations and Pairwise Frequencies) let you see how much the deletion method moves each estimate.

Assumptions and notes

EM and Little's test treat the quantitative variables as multivariate normal; strongly skewed variables can make the MCAR test reject too often.
EM estimates are valid when data are missing at random (MAR) given the variables in the analysis; they cannot fix data that are missing not at random.
Little's MCAR test needs at least two missing-value patterns; cases missing every quantitative variable are left out of EM.
With a sampling weight, only univariate statistics and patterns are reported; t tests, EM and Little's test assume independent cases and are withheld.

How the results are validated

The MVA engine is checked against independent Python references: univariate, listwise and pairwise statistics match numpy/pandas to 1e-11 relative, separate-variance t tests match scipy.stats.ttest_ind (Welch) to 1e-10, EM with the SPSS n − 1 denominator matches an independent numpy implementation to 1e-9 and Little's test to 1e-8, and maximum-likelihood EM with Little's test matches R norm::em.norm through naniar::mcar_test to 1e-7. Frequency weights are tested against the expanded data. Validation evidence

Questions

Little's MCAR test is significant. What now?

The data are probably not missing completely at random, so listwise deletion may bias results. Look at the separate-variance t tests to see which observed variables predict missingness, include them in the model, and use EM or Multiple Imputation, which assume MAR rather than MCAR.

Does a non-significant MCAR test prove the data are MCAR?

No. It only means there is no evidence against MCAR in these variables; with few cases the test has little power, and it cannot detect missingness that depends on the missing values themselves.

Can I use the EM means as my final results?

They are good point estimates under MAR, but they come without standard errors that reflect the missing information. For tests and confidence intervals, impute several datasets and pool them with Rubin's rules.

Are user-missing codes such as 99 treated as missing?

Yes. Codes declared as user-missing in the variable metadata are counted as missing, and a footnote lists how many there were per variable.

Is my data uploaded?

No. MVA runs in a web worker inside your browser tab; the file is never sent to Crosstabs servers.

Run Missing Value Analysis on your own data

Open a .sav, CSV or Excel file. Everything runs in your browser; the file never leaves your device.

Open the workspace

SPSS is a trademark of IBM Corp. Crosstabs is not affiliated with or endorsed by IBM.