LabKick Stats
Paste your data, say what you are comparing, and get the correct test with its justification and a publication-ready figure, in one act rather than three.
Top features
- The test is chosen and justified
“Welch's t-test. Your two groups have unequal variance; n = 3 so I checked normality, it is borderline, here is the non-parametric result too.”
- Stats and figure are one act
Not run a test, then build a bar chart, then place significance bars. One request returns the result and the figure, annotations already placed.
- Assumptions are always stated
Every test reports whether its assumptions hold, plus effect size, confidence interval and the exact p, not stars.
- It warns rather than flatters
“n = 3 is underpowered to claim this.” “You are running 14 comparisons, corrected for multiplicity here.”
- Prism's ceiling is intact
Mixed-effects and nested models, survival, curve fitting. “Fit a four-parameter logistic with shared top and bottom across datasets” works.
- A methods-ready report
Test, assumptions, effect size, CI and exact p, reproducible and citable, drop-in for a paper.
Everything in Stats
- t-tests, paired and unpaired
- One-, two- and multi-way ANOVA
- Post-hoc comparisons
- Non-parametric tests
- Correlation
- Linear and non-linear regression
- Survival / Kaplan-Meier
- Mixed-effects models
- Nested models
- Multiplicity correction
- Normality and variance checks
- Dose-response and EC50 fitting
- Enzyme kinetics
- Standard curves
- Custom equations
- Bar, scatter, box and violin plots
- Line and survival plots
- Error bars and significance annotations
- Styling in language
- Colorblind-safe and journal formats
- Direct manipulation for nudging
- Vector and raster export
- Explicit parameters and menus for power users
- Exact p, effect size and CI in every report
What changes
- Prism is a configuration tool. This is a “tell me what you found” tool.
- Choosing and configuring the test is the entire reason academic stats is hard and error-prone, and it is exactly the part the AI owns.
- A wrong p-value gets published, so the AI calls vetted statistical routines. It never computes a p-value in its head, and it would rather warn you than hand you a clean-looking lie.
Run it on your own data.
Stats is not sold on its own. One price gets you all nine tools, and the trial turns every one of them on.
No card to start · One login covers all nine tools · Cancel anytime