Analyze and verifyLimits and conditions

Analysis and verification:limits and conditions.

What a verified analysis establishes, the assumptions it carries, and the results SciGround will not issue.

What verification does not establish

Scientific truth: A verified result establishes specified properties of the calculation, the data connection, and the claim under stated assumptions. It does not prove the biology.
Assumptions from data shape: Normality, independence, and equal variances are not proven from the data. A non-significant diagnostic test is not evidence that an assumption holds.
Probability that a claim is true: Error rates describe the procedure across repeated use. They are not the probability that a particular conclusion is correct.

Borderline numbers: If the numerical bound cannot separate a result from its threshold, no significant or non-significant decision is issued; the result is unresolved.
No switching methods after the fact: SciGround does not switch methods until a result becomes significant. An alternative method is suggested only with an explanation of how it changes the question, and approval alone cannot turn a failed check into a guarantee.
Unsupported analyses: Analyses outside the supported workflows are refused rather than approximated.

Approximate methods: Some methods, such as Welch’s t-test, rely on statistical approximations. The numerical guarantee and the approximation are reported separately; no exact finite-sample error rate is claimed.
Sample-removal stability: Available for independent two-group Student and Welch comparisons with up to 12 samples per group, removing up to two. It is a statement about the conclusion under removals, not a measure of statistical validity, coverage, or false-discovery control. Preprocessing or selection not in the record is outside it.
AI-written text: See Use with your AI: only results that pass through SciGround are verified.