A test of whether your missing data is missing at random. What the three verdicts mean, the possibility the test can never rule out, and why that limitation matters more than the result.
MCAR, missing completely at random. Whether a value is missing has nothing to do with anything, observed or unobserved. A page was skipped by accident, a file was corrupted. Under MCAR, dropping incomplete cases loses power but does not bias your estimates.
MAR, missing at random. Missingness depends on things you did observe. Younger respondents skipped the income question more often, and you recorded age. This is the assumption most modern missing-data methods, including multiple imputation, are built on.
MNAR, missing not at random. Missingness depends on the unobserved value itself. The highest earners declined to state their income, and you have no way to know that from the data you have.
The distinction matters because it decides what you can legitimately do. MCAR permits listwise deletion. MAR calls for imputation. MNAR cannot be solved statistically at all and requires an argument about your design.
Little's test evaluates whether the pattern of missingness across variables is consistent with MCAR. Quanta reports one of three outcomes in plain language.
Missingness looks completely random. The test did not detect a departure from MCAR. Listwise deletion is defensible, though you still lose power.
Missingness likely depends on the data. A departure was detected. Complete-case analysis may bias your results, and imputation or a model-based approach is the appropriate response.
Not estimable. The test could not be computed, usually because there is too little missing data, too few variables, or a pattern too sparse to support the calculation. This is not a pass. It means the question is unanswered.
Note the direction, which works the same way as a normality test: a non-significant result is the reassuring one.
It cannot distinguish MAR from MNAR. This is the limitation that matters most. Little's test compares MCAR against everything else. Failing to reject tells you nothing about which of the two remaining mechanisms you are in, and there is no statistical test that can, because MNAR depends on values you do not have.
That is not a shortcoming of this implementation. It is a logical boundary. Deciding between MAR and MNAR is an argument from your design and your knowledge of the subject, not a result you can compute.
It has limited power at small samples. Failing to reject MCAR in a small dataset may mean the mechanism is benign or simply that the test could not tell.
It says nothing about who is missing. A dataset can pass Little's test while a particular group is systematically absent. That is a different question, and it is what the representation checks answer.
Open Data Quality. Little's MCAR sits with the missing-data diagnostics, alongside per-variable missingness and the missing-data patterns view.
Look at the patterns before the test. Which variables are missing together is often more informative than a single omnibus p value, because it points at the mechanism directly.
Missing-data diagnostics run on numeric fields, so import your numeric columns before expecting a result.
Use the action bar to add the result to a report, copy the table, or copy the APA text.
Report the test, the verdict, and the handling decision it led to.
Little's MCAR test was non-significant, χ²(42) = 38.71, p = .615, consistent with data missing completely at random. Complete-case analysis was therefore used (n = 238 of 250).
Report the amount and location of missing data alongside the test. A mechanism verdict without the quantity is not interpretable.
Say what you did and why. The test is only useful in a write-up as the justification for a handling decision.
Acknowledge that MNAR cannot be excluded. One clause is enough, and its absence is conspicuous to a methods reviewer.
Every result carries a ReliCheck Intelligence card with one action: explain this result in plain language. It runs entirely on your Mac with Apple Intelligence. No upload, no API key, nothing to configure. The model only explains the result Quanta has already computed and validated. It never computes a statistic, and it never applies a decision on your behalf.
From the action bar you can add the result to a report, copy the table, or copy the APA text. Exported reports carry the APA table, its note, and a caution specific to that analysis beneath it.
Quanta's multiple imputation is validated against R's mice package with Rubin's rules, matching pooled estimates, variance, and fraction of missing information. The engine is pure Swift with no R or Python at runtime. Full record on the validation page.
Data Quality is part of every Quanta subscription, $9.99 a month or $79.99 a year after the 30-day trial. One price for everyone. See pricing.
The analysis, the report builder, and the on-device explanations all work with the network off. Your dataset is never uploaded.
Multiple imputation (MICE) what to do when data is not MCAR · Representation checks who the missing data belongs to · Descriptive statistics per-field missing counts · the complete list on the analyses page.