Every feature from all five editions for 15 days, no sign-up.
Method Validation edition from US$ 475 a year · 30-day money-back guarantee.
Excel’s LINEST function and chart trendline fit an ordinary least-squares line and stop. The reference method is assumed error-free; there is no Deming or Passing-Bablok regression, no Bland-Altman plot, no bias at a decision point. Method comparison is not a single analysis. It is a sequence of decisions, and each answer changes which regression is valid and what the bias estimate means.
Is precision constant, or does it vary across the measuring range? Are there enough samples for a non-parametric method, or is a parametric one more appropriate? Is the bias constant or proportional? Does the linear model hold? And once the bias is estimated, does it matter at the concentrations where clinical decisions are made?
Which regression is valid depends on the data, so all five are available in the same analysis. Passing-Bablok, in the 1983 and extended 1988 forms, is the non-parametric starting point: robust to outliers, no distributional assumptions, no need to know the precision ratio. Deming and Weighted Deming use the known precision of both methods, constant SD or constant CV, for a more efficient estimate; OLS and WLS assume an error-free reference. Every fit reports slope and intercept with confidence intervals, systematic error split into constant and proportional bias, and Syx to flag matrix effects against expected within-run precision.
Regression estimates the bias; the difference plot shows how the differences are distributed across the whole range. Mean, median or linear fit bias models, with constant-width limits or V-shaped limits that widen when precision is proportional to concentration. Confidence intervals on the limits themselves, not only on the bias. The mountain plot alongside shows the cumulative distribution of the differences, a second view of agreement.
The overall slope gives the average bias, but clinical decisions happen at specific concentrations. Predict the bias, with a confidence interval, at each decision point you specify. Test equality (is there a significant difference?) and equivalence (is the difference clinically acceptable?) per EP09-A3, with the allowable difference as an absolute value, a percentage or both. A method can pass on bias and still fail when bias and imprecision together exceed the allowable total error. EP21-A adds the test method’s imprecision to the regression bias at each decision point and compares the total against allowable total error: one pass/fail assessment for both.
Precision that changes across the measuring range, or a material that does not behave like a patient sample, invalidates a single regression. Measure in singlicate, duplicate or replicate, with precision (SD or CV) and a precision plot for each method. Reduce the measuring interval, or partition it so each interval has its own regression, bias estimates and comparability assessment, as EP09-A3 specifies for non-constant precision. Per EP14-A3, compare each processed sample — calibrator or QC material — against the prediction interval from the patient sample regression; outside it, the material is non-commutable.
A regression is only as good as the linear model behind it. The CUSUM test, with exact p-values, and the Kolmogorov-Smirnov test check linearity before you trust the fit, and the CUSUM plot accompanies the test. Residual plots, raw and standardised, with a histogram show the scatter about the line; Pearson r summarises the correlation.
Not every comparison is of quantitative measurements. For qualitative tests — positive/negative, reactive/non-reactive — the proportions in positive and negative agreement quantify concordance between methods, with Clopper-Pearson exact and Wilson score confidence intervals. Kappa and weighted kappa adjust for chance agreement, with a test for agreement.
See method comparison results in detail — regression fits, difference plots, bias at decision points, and total analytical error — using CLSI example datasets you can download and follow along with.
EP09-A3 — Appendix I
EP09-A3 — Appendix IMethod comparison is one part of the Method Validation Edition, alongside measurement system analysis, reference intervals, and diagnostic performance.
Related guides in our Learn section: choosing a regression, Bland–Altman limits of agreement, bias at a medical decision point, how many samples you need, total analytical error, commutability, allowable total error, total error or measurement uncertainty, and comparing instruments within one laboratory.
Try it on your own data first. The 15-day trial is every feature from all five editions, with no sign-up and no licence key — install it and start straight away.
Method Validation edition from US$ 475 per year, or US$ 1155 for a perpetual licence. Every purchase carries a 30-day money-back guarantee. Need a quote for purchasing? Add the licence to the cart and save it as a PDF quote.
