Research Lab
The project workflow: proposal, data extraction, validation, statistics and results.
The project workflow
A Research Lab project keeps everything about one study together, under a single set of tabs: Overview, AI Recommendations, Proposal, Data Extraction, Statistics, Results, Manuscript and Notes.
- Proposal
- State the question and design. OrbitLab proposes the variables to collect and the analyses that fit.
- Data Extraction
- Build the extraction sheet, by hand or from an imported spreadsheet, and enter or import your data.
- Statistics
- Run the recommended analyses. Each result is saved as a card with its parameters and a timestamp.
- Results
- Review computed results, re-run them when the sheet changes, and export them.
Importing data
Comma, semicolon and tab separated files are supported, including files with unusual line endings. Numbers written with thousands separators are flagged rather than silently misread, and missing values are counted consistently across the app.
Wide spreadsheets with many columns are handled: from 0.1.5-beta a large or truncated response while building the extraction sheet is recovered rather than failing the import.
Validate Sheet and conflicts
Validation runs before analysis so problems surface while they are still cheap to fix. It checks types, missing values and whether the variables your proposal names are actually present in the sheet.
From 0.1.7-beta, a plan variable that already exists in your sheet under a shorter name is recognised instead of being reported as missing, so Validate Sheet and Resolve Conflicts can be completed.
What the statistics engine covers
Every inferential result is computed deterministically on your machine. No model sits in the computation path, so the same data produces the same number every time.
- Descriptive
- Summary statistics across your variables
- Parametric
- Welch's t-test, one-way ANOVA, Pearson correlation
- Rank-based
- Mann-Whitney U, Kruskal-Wallis, Spearman correlation
- Categorical
- 2×2 measures — odds ratio, risk ratio, risk difference, with confidence intervals
When the sample is smaller than planned
From 0.1.6-beta, collecting fewer participants than your proposal planned no longer blocks the analysis. It is reported as a limitation for you to mention in the write-up, and the target sample size is editable and is not overwritten when you validate the sheet.