I was asked this. I have calculated limits of agreement between two different methods of measuremen. Now I want to do this for men and for women separately. Is there a method I can use to assess the differences in limits of agreement between groups?
I thought this was interesting question. I think I would compare the mean differences and compare the standard deviations of the differences separately. For example, if you want a P value testing the null hypothesis that the limits of agreement are the same for men and women, you could do a two sample t test between the mean differences and an F test or Levene test between the variances of the differences. Then I would combine these two tests using a Bonferroni adjustment: multiply each P value by 2 and then choose the smaller one. This P value would then give a test of the null hypothesis that the limits are the same.
An alternative approach would be to compare the two upper limits of agreement, using the standard error of the limit, and the two lower limits using the same standard errors. We could find confidence intervals for each of these differences, or two P values which could be combined using Bonferroni, as above.
Back to frequently asked questions on the design and analysis of measurement studies.
Back to measurement studies menu.
Back to Martin Bland's Home Page.
This page maintained by Martin Bland.
Last updated: 18 June, 2015.
Back to top.