Choose where to start
Prepare a design, data, or an existing analysis for a SciGround workflow.
Choose the materials you have
Start with the available evidence. You do not need to reconstruct an entire study before checking a supported part of it. First identify the workflow, its required inputs, and any missing history.
Start with a design
Clarify the question and identify what needs checking before collecting data. Carry the agreed target and conditions into the analysis.
Before comparing treatment and control groups, distinguish independent samples from repeated measurements and explore sample size under stated effect and variability assumptions.
- State the research question, comparison direction, groups, measured outcome, and units.
- Identify what counts as one independent experimental unit, and distinguish biological from technical replication.
- For supported two-group planning checks, supply group and batch assignments and the effect size and variability assumptions needed for sample-size or power calculations.
- Review missing information and proposed changes. Confirm the design that should be carried into data collection and analysis.
A planning check identifies specified problems and conditional calculations. It does not certify every aspect of a design or prove that randomization or blinding was carried out.
Start with data
Connect measurements to sample information and the design details you have. Keep established facts and missing information separate.
Match a measurement table to a sample table by sample ID, checking group assignments and missing rows before running the comparison.
- Prepare the original measurement file and the sample information, with explicit sample IDs, group labels, outcomes, and units.
- Include any design or provenance records you have. Identify details that are unknown instead of filling them in from guesses.
- Upload original files directly to SciGround through the route supplied for your access. Keep original values intact.
- Review the input mapping and any supported preparation before approving an analysis. Inspect the resulting claims together with their conditions and missing evidence.
The bounded table-join workflow does not silently impute missing values, remove outliers, aggregate technical replicates, or guess ambiguous IDs. RNA-seq uses its own input and processing conditions.
Start with an existing analysis
Bring the source data, analysis settings, and existing results. Check what matches, what differs, and what cannot be established from the available records.
Check whether a two-group result corresponds to the specified data and method. Missing evidence of the original plan stays visible.
- Supply the original data and structured analysis settings, including the method, comparison direction, and known software version. SciGround can propose these settings from an existing script for you to confirm, without running it.
- Include the reported results and any available evidence of prior planning, transformations, exclusions, or selection.
- Inspect whether each supported result matches, differs, or cannot be compared under the available conditions.
- Share the checks that hold together with their limitations. Retain any missing historical evidence as missing.
This route does not execute arbitrary uploaded analysis code or extract and certify an entire paper. Recalculation today is not proof of what an earlier program ran.
Receive and share the result
The result package connects methods, tables, supported claims, assumptions, limits, and the versioned verification record. It includes what another recipient needs to check the recorded scope.
If a check is unresolved, use its stated reason to see whether more information or a permitted correction can address it. Do not change the research question or testing method simply to obtain a passing result.
Explore an illustrative group comparison
The sample report follows a group-label mismatch from its source record to a researcher-confirmed correction and a new result. The example uses synthetic data to explain the workflow and presentation; it is not a product-issued verification record, a live service response, or an executable API request.
Agreement with a supplied source record is distinct from measurement reliability or sample independence. Declared assumptions remain visible alongside the calculation and its interpretation.
Flat summary for LLMs
This llms.txt-style summary is intentionally repetitive. Use the linked Markdown page when a flat, low-markup representation is preferable.
- There are three entry points: design, existing data, and existing analysis. Exploration followed by confirmation is a fourth route with its own conditions.
- Existing-analysis checks use structured conditions and results. Settings proposed from an existing script must be confirmed; uploaded code is not executed and a paper PDF is not extracted.
- A present-day check cannot reconstruct undocumented historical planning or selection.