Bland-Altman Plot

Use Bland-Altman plot to inspect agreement between two measurements or methods, and keep that purpose distinct from simple correlation viewing.

Figure purpose

Bland-Altman plot is for agreement review, not for simple association review. It helps you inspect how two measurements differ across their shared range by plotting mean versus difference.

This is useful in method comparison, instrument comparison, and repeated measurement contexts where agreement matters more than correlation alone.

What the figure shows

The current figure contract displays the core agreement elements needed for a basic Bland-Altman read:

  • One point per paired observation, plotted as mean of the two measurements on the x-axis and difference between the two measurements on the y-axis
  • A central bias line showing the average difference between the two methods
  • Upper and lower agreement-limit lines showing the expected spread of differences around that average

This means the figure is designed to answer questions such as: "Is one method systematically higher or lower?" and "How wide is the disagreement across the measurement range?"

When to use or avoid

Use Bland-Altman plot when the question is "Do these two measurements agree closely enough?" rather than "Are they correlated?"

Avoid substituting a scatter or correlation-style figure when bias and spread of disagreement are the real concern. Two methods can correlate well and still disagree in a way that matters.

The current product has dedicated support for this figure, including bias and agreement-limit display. Docs should still keep the scope to the current figure contract rather than implying a complete agreement-method suite.

Licklider cannot determine automatically whether the observed agreement limits are acceptable for your scientific, clinical, or operational use case. It also does not decide for you whether disagreement grows with measurement magnitude, whether a proportional-bias adjustment is needed, or whether the paired measurements are close enough to be used interchangeably.

These limits matter because a Bland-Altman plot can show the pattern of disagreement, but only you can judge whether that pattern is acceptable in context. A narrow-looking spread may still be unacceptable for a high-stakes assay, while a wider spread may be acceptable in an exploratory setting.

Required columns

  • Two numeric measurement columns that represent the two methods, instruments, or readings being compared
  • A comparison context where mean-versus-difference interpretation is meaningful

The current implementation is dedicated to this figure family rather than a generic scatter overlay.

Bland-Altman plot should be read differently from Scatter Plot and Regression Plot. Those figures focus on relationship patterns. Bland-Altman focuses on bias and agreement limits.

For current scope boundaries, keep Figure Support Matrix and Known Limitations nearby.

Design Rationale & References

This page follows a simple rule: method-comparison figures should emphasize disagreement directly, not leave the user to infer agreement from correlation. That is why the figure is framed around mean-versus-difference, with explicit bias and agreement-limit display, instead of reusing a generic scatter-only contract.

The distinction from correlation views is intentional. Correlation can remain high even when two methods disagree by a clinically or scientifically important amount. A Bland-Altman display makes the average bias and the spread of differences visible in the same view, which is the core question when deciding whether two measurements agree closely enough to substitute for one another [1, 2].

The scope is also kept narrow on purpose. The current figure contract supports the core bias-and-limits reading, but it is not presented as a complete agreement-analysis suite. That boundary helps avoid over-promising advanced agreement diagnostics that are not part of the current product surface.

  1. Bland, J. M., & Altman, D. G. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. The Lancet, 327(8476), 307-310. https://doi.org/10.1016/S0140-6736(86)90837-8
  2. Giavarina, D. (2015). Understanding Bland Altman analysis. Biochemia Medica, 25(2), 141-151. https://doi.org/10.11613/BM.2015.015

Alternative figures

  • Use Scatter Plot when the main question is relationship shape rather than agreement.
  • Use Regression Plot when a fitted trend is the main goal.
  • Keep Bland-Altman as the stronger choice when method comparison or instrument agreement is the real question.

TODO (Phase02+)

  • Expand only if the docs later cover more advanced agreement-analysis variants beyond the current bias-and-limits figure contract.