Limits of agreement - should I use ANOVA MS?

[lucy smith]

Hi all,

I am trying to calculate 95% limits of agreement, rather than using bias +/- 1.96 SD I have read that using bias +/- 1.96 (Squareroot of 2 x mean square error from ANOVA) is a better approach as this accounts for heteroscedasticity, would i be right in assuming this and just to double check the mean square errors are the "mean square sphericity assumed" under the within-subjects output of SPSS?

Any help would be hugely appreciated!

Thanks in advance!

[Bernard]

Re: Limits of agreement - should I use ANOVA MS?

Hi,

You can also use the variance components approach below, which may have some advantages e.g. imbalanced study design, ability to model heteroscedasticity in the mixed model. Sample codes provided in the paper.

Int J Biostat. 2008;4(1):Article 16.
Statistical models for assessing agreement in method comparison studies with replicate measurements.

Carstensen B1, Simpson J, Gurrin LC.

Bernard

[lucy smith]

Re: Limits of agreement - should I use ANOVA MS?

Thank you for this response it is extremely helpful - very grateful!