Guides · BioStudioAI
Choosing a statistical test
Choose the test from the design and outcome type - paired or independent, number of groups, distribution and sample size - before seeing results. Report effect size and uncertainty alongside any p-value.
Worked example
Take a concrete case and work it with the matching calculator in Tools rather than by hand, so the arithmetic is checkable. Record the inputs, the formula and the result together; the calculator pages on this hub show each step with your numbers, which is the same discipline to apply in a notebook or report. When the example involves a curve, check back-calculated standards or residuals before quoting the sample result, and state the units at every step.
Common mistakes
| Mistake | Why it changes the answer |
|---|---|
| Reading samples outside the validated range | Small signal errors become large concentration errors |
| Testing on technical replicates as if biological | Sample size and significance are overstated |
| Choosing the analysis after seeing the data | Any threshold can be made to look favourable |
| Reporting a summary without the individual points | Failed replicates and bimodal data stay hidden |
FAQ
| Question | Answer |
|---|---|
| What should I report with the result? | The method and version, controls, acceptance criteria, exclusions with reasons, and the raw file identifier. |
| Where do I calculate it? | Use the matching free calculator in Tools; it shows its working. |
Link up to the topic hub for this assay family, run the matching Protocol for the bench method, and see Troubleshooting when a run fails these criteria. Have plates of this data? BioStudioAI fits, quality-checks and reports them with an audit trail: see /platform/.
Frequently asked questions
What should I report?
Method, controls, acceptance criteria, exclusions and the raw file.
References
- Assay Guidance Manual · NCBI Bookshelf (2024)
- MIQE Guidelines · Clinical Chemistry (2009)
- ICH M10 Bioanalytical Method Validation · ICH (2022)
Upload the raw file and BioStudioAI fits, quality-checks and reports it with an audit trail.
See the platform →Related
ANOVA asks whether group means differ more than within-group scatter can explain, then post-hoc tests locate the differences with correction. It assum
Use Bonferroni or Holm when a single false claim is costly (confirmatory work), and Benjamini-Hochberg FDR when screening many candidates for follow-u
ALCOA+ asks that records be attributable, legible, contemporaneous, original, accurate - and complete, consistent, enduring and available. It is a hab