One rater assess twice the angle of bone formation. The units are continuous - degrees. n=100.
I know how to create a Bland-Altman plot for inter rater. But this is not the case. I tried to seek for information on internet and there are some papers on the topic, but I just did not get them. Can somebody suggest a simple way how to conduct this analysis of repeated measures in SPSS?
the procedure is very similar (actually the same) to what you did for inter-rater assessment of two repeated measurements by two different raters. since you refer to a "reproducibility analysis" I would suggest to use other methods and parameters as well (beyond Bland-Altman) such as: ICC: intra-class correlation coefficient and others.
I also suggest to take a look at this approach: J Hum Hypertens. 2012 Oct;26(10):594-602
Besides the statistical assessment of two repeated measurements there should be also a clinical or physiological evaluation of their absolute difference, in terms of clinical relevance.
Thanks Theodore. I'm already employing ICC. But ICC shows the strength of correlation. It says nothing about the differences in actual figures expressed in degrees. This is the reason I'm going to construct BA-plot. The procedures are similar (you are correct), but not the same as the variance should be different in inter- and inter-rater settings, as I understand. We have independent groups in one case and dependent in another. I suspect, there has to be a small difference in the way to calculate limits - maybe instead of employing SD of mean difference I should use an adjusted SD (Limit = Mean +- 1.96*SD). Is that so and what is the simplest way to adjust?
I apologize. I should write: "...the variance should be different in inter- and intra-rater settings..."
Hello, Mikhail.
In the assessment of reliability with quantitative data, it is suggested that no single estimate is appropriate, and that a combination of statistics provides a better picture of the true reliability.
In your case, ICC and Bland-Altman plot (I completely agree), but there are other couples of options, like Lin´s coefficient (CCC) and Bland-Altman, or the Standard Error of Measurement (SEM= SD∙√1-ICC) and Bland-Altman method.
They all have their pros and cons.
Have a look at:
- Badges
- Science topic
- Similar topics
- Papers