I run a regression for bland-altman analysis and generated the model D=a+bM, where D is the differences between the methods, and M is the mean of the methods.
I used these values to generate upper and lower limits of agreement, using the equation below
D±1.96*SD, where SD is the standard deviation.
I was able to compute the upper and lower limits of agreement, but am stuck on the confidence interval for these LOAs, I am mostly interested in the confidence intervals for the highest upper and lower LOA point and the lowest upper and lower LOA point, as shown in the image. Please note, the image is for reference purposes only.
I have tried to look up similar questions, but they only give the direct confidence intervals for the regression line, which I think in my case wouldn't work. Does anyone have a suggestion for this? I work with R but a manual solution will also do. I should also be able to understand a code from another program.
Thanks
What is SD the standard deviation of? Once you know that then you know for sure what agreement means. The 1.96 says the you are 95% confident in your band. If i understand what you want to do this reference may be helpful:
: https://www.google.com/search?q=regression+agreement+limits&rlz=1C1CHBF_enUS847US847&oq=regression+agreement+limits&aqs=chrome..69i57j69i60l2.33566j0j8&sourceid=chrome&ie=UTF-8
Best, D. Booth
The SD is in this case the residual standard error from the regression
Maybe, it would help to do it in R. There, plotting regressions including confidence intervals is pretty easy: https://rstudio-pubs-static.s3.amazonaws.com/71339_d0b8346f41314979bc394448c5d60d86.html
In base R, you can use the predict function for predicting the regression lines and confidence intervals with a confidence level as needed. Or you can use ggplot2, where it is even easer to already include the standard errors in the plot without having to calculate regressions at all manually ( https://rpubs.com/aaronsc32/regression-confidence-prediction-intervals ).
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression