regressionsmodeller

Regressionsmodeller is a software framework for building, fitting, and evaluating regression models. It provides a unified interface to specify model families, choose estimation methods, and generate predictions from data. The toolbox covers traditional linear regression, generalized linear models, and non-linear variants, as well as robust and regularized approaches. It can operate on tabular datasets with numeric and categorical features, and emphasizes reproducibility and interpretability of results.

Core features include model specification language, automatic preprocessing (handling missing values, encoding categorical variables, scaling), regularization options (L1, L2, elastic net), and feature engineering helpers (polynomials, interactions). It supports cross-validation, train-test splits, and automated model selection using information criteria or cross-validated metrics. Diagnostic tools provide residual analysis, influence measures, and goodness-of-fit statistics. The framework also offers model export, scoring, and uncertainty estimation for predictions.

Typical workflow: ingest data, preprocess, specify candidate models, fit with chosen estimation method, evaluate on validation data, compare alternatives, and select a final model. It supports hyperparameter tuning, multi-model comparisons, and reproducible experiments via configuration files and provenance logs. Predictions can be produced for new data, with confidence intervals and prediction intervals when appropriate.

Regressionsmodeller is intended for data scientists, researchers, and analysts across industries, including economics, biology, engineering, and social sciences. It emphasizes interoperability with existing data pipelines and export formats, allowing models to be deployed in reporting dashboards, notebooks, or production systems.