Plan an experimentLimits and conditions

Design checks:limits and conditions.

What the pre-experiment check examines, what it assumes, and what remains your responsibility.

What it does not do

No “design is valid” verdict: Machine checks, assumption-dependent calculations, external assumptions, and expert judgment are reported separately. Passing the checks does not certify the whole design.
No automatic optimal design: SciGround compares the candidates you consider. It does not search for an optimal experimental design.
No proof of how the study was run: It cannot confirm that randomization or blinding actually happened, and it is not an ethics review.
No post-hoc power: Power computed from an observed effect is not used to justify a conclusion after the fact.

Assumed effect and variability: A sample size is only as good as the effect size and variability you assume. SciGround shows the assumptions; it does not know the true values.
Your declarations: Real-world independence, allocation, and measurement remain facts you report. SciGround records them as stated conditions, not as proven facts.
Locking is a timestamp, not proof of blindness: Locking records when you registered the plan. It cannot show that no outcome data had been collected or seen before then.
Dropout allowance: Enrolling extra subjects for expected dropout does not make later missing data analyzable.

Two independent groups: The design check targets studies with two independent groups and one continuous outcome. Designs with three or more groups, repeated measures, or multiple factors are not checked.
Two-stage plans: Exactly two pre-fixed looks, two independent groups, one outcome, and a two-sided Student’s t-test at 2.5% per look. The model assumes normal errors, one common variance across groups and stages, constant group means across stages, and complete planned data at each look. Adaptive sample sizes, futility stopping, three or more looks, and general alpha-spending are not supported; no sample-size saving is claimed, and it is not for clinical use.
Stops for other reasons: Deaths or welfare stops end a study operationally. They are recorded with their reason and are not treated as a statistical result.
Changes after data arrive: A design changed after data are received becomes a new version. It cannot be relabeled as the original pre-specified plan.