Talk:Bland–Altman plot

It would be good if some could explain how to interpret the graph. For example what is considered as a good agreement. Thanks. 194.83.140.121 (talk) 10:23, 8 February 2008 (UTC) 2nd that.

Should Tukey plot really redirect here, rather than vice versa? The article already acknowledges that Tukey has prior art, and that only one field of science refers to it as a Bland-Altman plot. 131.227.215.82 (talk) 12:06, 12 January 2011 (UTC) Answer: I know Tukey from Tukey Boxplot ... Bland Altman Plot seems to be something different.

Range of application
The controversy should be better explained.

The title of the original article clearly indicates "two clinical measurements". The authors later claim that this method can also apply to comparison with a gold measure, and that's true. However that not indicates that it is the most appropriate. This method is not directly intended to compare a new method to a gold one, but gives a statistical index of the similarity-dissimilarity that can help to judge if an observed difference between the methods is possibly obtained by chance. By application, this method is then clearly appropriate or directed to clinician who must know how much different from the norm a value should be to indicate with uncertainty a pathology.

Special methods based on regression are generally preferred when comparing to gold standard, because they allow to clearly and visually distinguish: - the relationship between new and gold method. - the dispersion of the data around the relationship (the error of the new method compare to the gold measure). - the evolution of this dispersion into the the range of values when errors are non-uniform (a very common case in biological data). —Preceding unsigned comment added by 92.49.68.223 (talk) 12:56, 19 February 2008 (UTC)

respectfully
I work in biotech, and know how hard it is to explain stuff like this, and would like to make some suggestions Avoid words like "abcissa", unless you have reason to think that people who don't know what an x axis is will know what an abcissa is (ie, if someone knows what an abcissa is, they will know what an x axis; however, someone who knows what an X (horizontal) axis ismay be unfamiliar with abcissa Give a more specific example; objects of volume is not specific; say height of 10 children in a class, and then give a table with all the values, so people can see, clearly, how measured values are plotted - I'm sorry but I think there is some ambiguity in the phrasing I personally don't contribute anymore to wiki because of this: the copyleft allows *for profits* to resell our work; I don't mind someone reposting my hard work; I find it unacceptable that someone else will make money off of myhard work; ymmv — Preceding unsigned comment added by 68.236.121.54 (talk) 16:15, 8 December 2011 (UTC)

'agreement' vs 'correlation'
"A high correlation for any two methods designed to measure the same property is thus in itself just a sign that one has chosen a wide spread sample. A high correlation does not automatically imply that there is good agreement between the two methods." So what is good agreement? Can there be agreement without correlation? Something is missing from this section. --98.70.128.80 (talk) 21:57, 13 May 2012 (UTC)

Tukey?
The article asserts without reference that Bland and Altman's plot is "identical" to a plot by Tukey; I haven't been able to find a reference for that, but I have been able to find an article in PLOS One which makes that assertion - citing this page! Can anyone, perhaps from an engineering background, help find a reference to Tukey's plot? 163.1.200.252 (talk) 10:19, 4 August 2016 (UTC)Richard Stevens

Tukey Mean-Difference Plot is described here: https://www.itl.nist.gov/div898/software/dataplot/refman1/auxillar/meandiff.htm. But these graphs seem to be different. It uses difference vs means, but in a different context, not comparing measurements. So it looks like the "Tukey" claim is misplaced?

Dhanish