Nine practical utilities for statistical analysis, machine learning evaluation and business planning — rigorous calculations with accessible interfaces, no formula memorisation required.
Whilst demonstrations showcase possibilities, tools solve immediate practical problems that practitioners encounter daily. Each calculator combines rigorous statistical foundations with accessible interfaces, making best practices the path of least resistance rather than a barrier requiring specialist knowledge.
Three tools that make rigorous analysis the path of least resistance — turning parameter selection, power evaluation and A/B interpretation from barriers into routine steps.
Adapts to A/B tests, survey design and observational studies. Parameter tooltips guide non-statisticians through every input.
Evaluates completed studies to distinguish genuine null results from underpowered ones — post-hoc analysis most practitioners skip.
Comprehensive analysis covering design validity, statistical inference and business impact in one workflow.
| Preset Profile | Parameter Guidance |
|---|---|
| Exploratory research | More lenient parameters appropriate for early-stage investigation. |
| Regulatory submission | Stringent controls to satisfy compliance requirements. |
| Growth experiment | Balances speed against rigour for fast-moving product teams. |
Common evaluation tasks that interrupt workflow when the code isn't handy — instant calculations and visualisations that maintain statistical rigour without custom scripting.
Transforms a raw confusion matrix into a full diagnostic suite aligned with actual business or safety objectives.
Publication-quality ROC curves with multi-model comparison and principled operating-point selection for any decision context.
Multiple importance methods revealing different facets of model behaviour — plus honest guidance about what the numbers do and do not mean.
| Criterion | How it selects a threshold |
|---|---|
| Youden index | Maximises the sum of sensitivity and specificity — appropriate when both error types matter equally. |
| Cost curve analysis | Minimises total costs when you specify relative costs of false positives and false negatives. |
| Clinical utility | Identifies thresholds based on prevalence and minimum acceptable sensitivity or specificity for the application. |
Paste prediction scores and labels below — the curve, AUC and optimal threshold recompute instantly, rendered locally with d3.
No data yet — load the sample or paste your own scores and labels above.
Positives: — · Negatives: —
Translate technical outputs into the financial and operational language stakeholders actually care about — connecting model performance to organisational decisions.
Systematic ROI estimation that avoids both under-valuing data science and losing credibility through over-optimism.
Realistic estimates built from historical project data, with dependency mapping and resourcing scenario comparison.
Right-size teams based on project portfolios — balancing workload, skill requirements and budget rather than simply maximising headcount.
Adjust the value drivers and costs below — ROI, NPV and payback recompute instantly, with a Plotly sensitivity chart rendered locally.
Net: —
Key Insight — Practical tools make rigorous statistical analysis accessible without memorising formulas, and bridge technical model outputs with the organisational value stakeholders actually care about.
These tools are fully interactive in the DataScienceDNA app. Explore the Lab →