Validity of linear regression in method comparison studies: is it limited by the statistical model or the quality of the analytical input data?

Abstract

We compared the application of ordinary linear regression, Deming regression, standardized principal component analysis, and Passing–Bablok regression to real-life method comparison studies to investigate whether the statistical model of regression or the analytical input data have more influence on the validity of the regression estimates. We took measurements of serum potassium as an example for comparisons that cover a narrow data range and measurements of serum estradiol-17β as an example for comparisons that cover a wide data range. We demonstrate that, in practice, it is not the statistical model but the quality of the analytical input data that is crucial for interpretation of method comparison studies. We show the usefulness of ordinary linear regression, in particular, because it gives a better estimate of the standard deviation of the residuals than the other procedures. The latter is important for distinguishing whether the observed spread across the regression line is caused by the analytical imprecision alone or whether sample-related effects also contribute. We further demonstrate the usefulness of linear correlation analysis as a first screening test for the validity of linear regression data. When ordinary linear regression (in combination with correlation analysis) gives poor estimates, we recommend investigating the analytical reason for the poor performance instead of assuming that other linear regression procedures add substantial value to the interpretation of the study. This investigation should address whether (a) the x and y data are linearly related; (b) the total analytical imprecision (sa,tot) is responsible for the poor correlation; (c) sample-related effects are present (standard deviation of the residuals ≫ sa,tot); (d) the samples are adequately distributed over the investigated range; and (e) the number of samples used for the comparison is adequate.

Comparative clinical study of canine and feline total blood cell count results with seven in‐clinic and two commercial laboratory hematology analyzers

“…Results were statistically analyzed by linear regression, Deming regression (normally distributed measurement errors), Passing-Bablok regression (nonnormally distributed measurement errors) and Bland-Altman plots using Microsoft Excel with Analyse-it (Analyseit Software Ltd., Leeds, UK) and GraphPadPrism version 4.0 (GraphPad Software Inc., La Jolla, CA, USA). 18,19 Results for MCH, MCHC, red cell distribution width, and nucleated RBCs were not evaluated in this study.…”

Automated flow cytometric cell count and differentiation of canine cerebrospinal fluid cells using the ADVIA 2120

“…The intercept and slope must be interpreted with caution, as r-values were o .975. 12 The sample with 4 50% difference between ADVIA 2120 and manual differential results had a low TNCC of 5/mL. Seven of 21 samples with moderate to marked pleocytosis (TNCC 30-1747/mL) due to elevated lymphocyte counts were correctly classified by the ADVIA 2120.…”

Respiratory snorkel and valve system for breath‐by‐breath gas analysis in swimming

“…Before statistical analysis, individual B Â B data were averaged every 60 s for exact time matching and B Â B noise (inter-breath fluctuation) reduction using the filtering data tool of the manufacturer's software as well as to simplify the data before further statistical analyses. Linear correlation analysis between averaged, pooled data of both tests (using CM and SV, respectively) was used as a first screening test for the validity of linear regression and data agreement (Stockl et al, 1998), and the Pearson's coefficient of determination (R 2 ) was computed. In addition, agreement between both methods was evaluated for each parameter investigated by Passing-Bablok regression analysis of the pooled data (Passing & Bablok, 1983, 1984.…”