Descriptive and inferential
Distributions, tests, intervals, power, equivalence and non-inferiority.
Analyze
The result does not begin with a p value.
Eikos puts estimation, uncertainty and method validity before a significance label.
A traceable workflow
- Define the question and variables.
- Inspect quality, missingness, scales and design.
- Review compatible methods and their rationale.
- Check assumptions and robust alternatives.
- Read estimates, intervals, effects and diagnostics.
- Export conclusions and reproducible code.
Refusal is also a result. If the data cannot support the requested inference, Eikos does not manufacture a number.
Missing data
Missingness patterns, complete-case cost, multiple imputation, Rubin rules and declared-δ MNAR sensitivity.
Causal design
DAGs, adjustment sets and instrumental variables with 2SLS, strength diagnostics and Anderson–Rubin.
Modelling and synthesis
Linear, generalized and mixed models, survival, time series, surveys, meta-analysis and heterogeneity.
Machine learning
Classification, regression, ensembles, SVMs, validation and explainability.
External checks
Reference values generated with R and Python, with figures checked in Chromium, Firefox and WebKit.