pubmed.info
Drag tiles by their title, resize at the bottom edge.

Bland-Altman plot generator – bias and limits of agreement

Paste paired measurements from two methods and get the Bland-Altman plot with bias, limits of agreement and their confidence intervals, plus checks and a suggested text. Free, no account; your data are not uploaded.

More about this tool

What you can do

  • Compare two measurement methods: bias (mean difference) and limits of agreement (bias ± 1.96 SD), each with a confidence interval at 90, 95 or 99%.
  • Analyse absolute differences (A − B), percentage differences or logarithmic differences (ratio A / B, also back-transformed).
  • Plot the difference against the mean of both methods or against reference method B.
  • Check the assumptions: Shapiro-Wilk test, skewness and kurtosis of the differences, regression of the difference on the mean (proportional error) and of the absolute differences (spread). The tile points out when the limits may be misleading.
  • See the t test of the bias and how many pairs lie outside the limits.
  • Enter a clinically acceptable difference for the suggested text.

How it works

  • One row per pair: value of method A, value of method B, optionally preceded by an identifier such as the patient. Separate by tab, semicolon or space, e.g. pasted from Excel or SPSS; decimal comma and point are both accepted and a header row is allowed.
  • 3 to 5,000 pairs. Rows with missing values are excluded.
  • Name the methods, set the unit, axis ranges and decimals, and optionally shade the confidence ranges or draw the regression line.

Export

SVG with real text and your input embedded, vector PDF (Latin script only) or PNG at 300 or 600 dpi, 75 to 170 mm wide, in colour or black and white. Key figures as CSV and a suggested text for methods and results. Load the SVG or a saved project file to continue later.

Your data: nothing is transmitted to a server. Only this browser remembers your input and settings.

Frequently asked questions

How do I report a Bland-Altman analysis?

Give the number of pairs, the bias and the limits of agreement, each with its confidence interval, and compare them with a clinically acceptable difference defined before measuring. The tile writes a suggested text with these figures that you can check and adapt.

When are two methods interchangeable?

Not on the basis of a P value: what matters is whether the limits of agreement are narrow enough for the clinical purpose. Judge this at the outer end of their confidence intervals, which are wide with fewer than about 50 pairs. A correlation coefficient alone says nothing about agreement.

Can I use repeated measurements per person?

No. Several measurements per person violate the assumption of independent pairs, and the method of this tile is not suitable for them.

pubmed.info sets no tracking cookies and builds no user profiles. Visits are counted anonymously (cookieless, shortened IP addresses). Ads from Google appear only according to your consent choice.