Bias-variance decomposition of absolute errors for diagnosing regression models of continuous data

Summary Bias-variance decomposition (BVD) is a powerful tool for understanding and improving data-driven models. It reveals sources of estimation errors. Existing literature has defined BVD for squared error but not absolute error, while absolute error is the more natural error metric and has shown advantages over squared error in many scientific fields. Here, I analytically derive the absolute-error BVD, empirically investigate its behaviors, and compare that with other error metrics. Different error metrics offer distinctly different perspectives. I find the commonly believed bias/variance trade-off under squared error is often absent under absolute error, and ensembles—a never hurt technique under squared error—could harm performance under absolute error. Compared with squared error, absolute-error BVD better promotes model traits reducing estimation residuals and better illustrates relative importance of different error sources. As data scientists pay increasing attention to uncertainty issues, the technique introduced here can be a useful addition to a data-driven modeler's toolset.

• Publication year

  2021

• Venue

  Patterns

• Publication date

  2021-07-21

• Fields of study

  Mathematics, Computer Science, Medicine

• Identifiers
  DOI 10.1016/j.patter.2021.100309PMID 34430928PMCID 8369249
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar, PubMed

• No claims are published for this paper.

• No concepts are published for this paper.

• Per-pixel bias-variance decomposition of continuous errors in data-driven geospatial modeling: A case study in environmental remote sensing

  2017cited by this paper
• Ambiguities inherent in sums-of-squares-based error statistics

  2009cited by this paper
• Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance

  2005cited by this paper
• Variance and Bias for General Loss Functions

  2003influential reference
• A Unifeid Bias-Variance Decomposition and its Applications

  2000cited by this paper
• Boosting the margin: A new explanation for the effectiveness of voting methods

  1997cited by this paper
• Bagging Predictors

  1996cited by this paper
• Bias, Variance , And Arcing Classifiers

  1996cited by this paper
• Neural Networks and the Bias/Variance Dilemma

  1992influential reference

• Optimizing the thermal performance of double skin Façades in a cold climate with phase change materials and cavity ventilation

  2026cites this paper
• Photogrammetric image processing and comparison of experimental result with laser diffraction and the PDA method

  2025cites this paper
• A new method of calculating crown projection area and its comparative accuracy with conventional calculations for asymmetric tree crowns

  2024cites this paper
• Sugarcane water requirement and yield projections in major producing regions of China under future climate scenarios

  2023cites this paper
• Ensemble Learning, Deep Learning-Based and Molecular Descriptor-Based Quantitative Structure–Activity Relationships

  2023cites this paper
• Decomposition of the mean absolute error (MAE) into systematic and unsystematic components

  2023cites this paper
• A Survey on Epistemic (Model) Uncertainty in Supervised Learning: Recent Advances and Applications

  2021cites this paper
• A discrete seasonal grey model with disturbance adjustments and dummy parameters for cyclical effects in electricity demand forecasting

  year unknowncites this paper