Paper · S7634
Bayesian evidence and model selection
Knuth and colleagues explain Bayesian evidence, prior-sensitive model comparison and approximate estimation. Their examples show how assumptions, dependent observations and numerical uncertainty affect apparent model winners. The paper provides methods context for evaluating claims; its sensor example overlaps an earlier study and supplies no independent UAP observations.
Read the paper at knuthlab.orgWhat we checked
Read all18 PDF text leaves/102663 extracted characters,110 numbered equations,134 reference entries,five footnotes and all captions. Viewed eight native page images covering all nine figures and both tables. A truncated text return was repaired with an explicit complete reread. Native available-online date is29June2015; teaching-folder2020 is not publication date. DOI10.1016/j.dsp.2015.06.012.
Complete supplied methods text and selected native figures/tables read
How we check sources
Not checked yet
- Ten other page layouts and unselected fine equation glyphs remain unviewed. Native equation sign/normalization conventions and a nonmonotone graph-axis label remain clarification gaps. Later errata were not checked.
- Scientific code, original measurements, cited bodies and numerical replication were not reviewed. Synthetic epochs, calibration orientations and computational replicas are dependent units. Model and prior assumptions limit comparisons; the highest mean score may not be a clear winner within uncertainty.
- The sensor example depends on the study already represented by S7575. Named university, DFG and NASA/JPL support and historical biographies do not report a secret government approach, recovered-object custody or current roles. Biological illustration is recorded only as inference context.
Cite this record
Kevin H. Knuth, Michael Habeck, Nabin K. Malakar, Asim M. Mubeen, Ben Placek. Bayesian evidence and model selection. June 29, 2015. BioParity S7634. https://bioparity.io/sources/S7634Record data (JSON)