Saved Search

Method Comparison Regression Analyzer

Estimate the relationship and fitted bias between two measurement methods using Deming and Passing–Bablok regression.

1. Paired measurements

2. Comparison setup

Analysis directionCandidate method (Y) regressed on Comparator method (X)

3. Regression methods
Regression methods

Use weighted Deming when measurement error grows in proportion to concentration. It requires strictly positive X and Y measurements and is fitted iteratively.

λ = σ²error,Y / σ²error,X. A value of 1 assumes equal analytical error variances. Use independent precision, replicate, or QC evidence where available.

4. Decision pointsOptional · none specified

Estimate fitted bias at up to 10 Method X values. Acceptance criteria are not supplied.

About this analysis

Example clinical scenario

A laboratory compares a candidate assay with a comparator using one measurement from each procedure on every independent specimen. The investigators estimate the fitted relationship and bias at prespecified medical decision concentrations.

Methods

Standard unweighted Deming regression uses an externally justified ratio of Y analytical error variance to X analytical error variance. Classical Passing–Bablok uses the exact 1983 tangent shifted-median convention. Deming intervals use leave-one-out jackknife standard errors with a Student t interval; Passing–Bablok coefficient intervals use analytical ranks, while decision-point bias intervals use a deterministic paired-row BCa bootstrap.

Assumptions and limitations

Pairs must be independent and measure the same or a scientifically comparable quantitative construct on a common scale. The tool supports unweighted Deming and Linnet constant-CV weighted Deming. It does not support arbitrary observation weights, replicate-aware, clustered, censored, nonlinear, or calibration models. No observation is deleted automatically.

Validation status

Version 1.1.3 completed automated numerical benchmarks, import and privacy checks, responsive browser review, and an independent Python reproduction of the deterministic BCa benchmark before catalogue release. Version 1.2.0 retains statistical method version RA04-M2 and completed its automated interface contract, rendered responsive, keyboard, accessibility-tree, and A4 print/PDF release checks on 8 September 2026. This evidence does not constitute clinical, regulatory, or universal statistical validation.

Current application version: 1.3.0. The September interface refinements follow the BA/ICC workflow and retain RA04-M2 calculations. The current layout awaits owner review, native PDF inspection, and screen-reader listening checks.

Privacy

All calculations and file parsing occur locally in your browser. Data are not intentionally stored or transmitted. Do not enter names, record numbers, dates of birth, or other identifiers. Spreadsheet formulas and macros are not executed.

Selected references

  1. Passing H, Bablok W. J Clin Chem Clin Biochem. 1983;21:709–720.
  2. Linnet K. Clinical Chemistry. 1993;39:424–432.
  3. NIST/SEMATECH Dataplot Reference Manual: Deming Fit.
  4. CLSI EP09. Measurement Procedure Comparison and Bias Estimation Using Patient Samples. Corrected 3rd ed.

Suggested citation

Yaacob, NM & Che Soh, NAA. Method Comparison Regression Analyzer, version 1.3.0. NajibMY; 2026. Available from: https://najibmy.com/tools/method-comparison-regression/

These analyses support research and do not establish agreement, interchangeability, clinical acceptability, or regulatory validation.