Statistical Methods for Diagnostic Evaluation¶
This page explains the statistical methods used to evaluate diagnostic tests — in plain language, with the interpretation each method supports.
ROC Analysis¶
What it does: Plots sensitivity against (1 − specificity) across all possible cutoffs; the AUC (area under the curve) summarizes overall accuracy.
| AUC | Interpretation |
|---|---|
| > 0.9 | Excellent |
| 0.8–0.9 | Good |
| 0.7–0.8 | Fair |
| < 0.7 | Poor |
ROC analysis also identifies the optimal cutoff — the threshold that best balances sensitivity and specificity for the clinical context.
Bland-Altman Plot¶
What it does: Plots the difference between two methods against their mean, visualizing bias (systematic difference) and limits of agreement.
- A mean difference near zero → no systematic bias.
- Narrow limits of agreement → the methods agree closely.
Passing-Bablok Regression¶
What it does: A non-parametric regression that estimates systematic bias (intercept ≠ 0) and proportional bias (slope ≠ 1) between two methods.
| Result | Meaning |
|---|---|
| Intercept ≠ 0 | Constant (systematic) bias |
| Slope ≠ 1 | Proportional bias (bias grows with concentration) |
Confidence Intervals (95%)¶
What they do: Express the precision of an estimate. A sensitivity of "92% (95% CI 85–96%)" means the true sensitivity likely falls in that range — a wide interval signals a small or noisy study.
FAQ¶
Why use both Bland-Altman and Passing-Bablok? They answer different questions — Bland-Altman visualizes agreement, Passing-Bablok quantifies systematic and proportional bias. Together they give a complete picture.
What is a good AUC for a veterinary diagnostic test? Above 0.9 is excellent; 0.8–0.9 is good. For screening tests, prioritize sensitivity; for confirmatory tests, specificity.
For study designs, see Study Design.
Authored by: Migibio Clinical & Scientific Affairs, Guangzhou Magic Biotech Co., Ltd.
Reviewed by: Migibio R&D Quality Committee
Last updated: 2026-08-13
Disclosure: Migibio (Guangzhou Magic Biotech Co., Ltd.) is the manufacturer of the FIA680/FIA880 analyzers and FICT reagents referenced in this content. See our Editorial & Review Policy.