Compare 2 or more devices

#1
Hi

I am planning to compare 2 (eventually 3) devices. I will perform 3 measurements at one time point with each device. So I will have 3 measurements with device A, 3 measurements with device B (and eventually 3 measurements with device C), all of them at the same timepoint in X subjects.

I will do a bland altman analysis to compare 2 devices, caclulate means and SD for each 3 measurements per device, precision (1.96xSD) and repeatibility index (=SDx2.77) and CoV (=SD/mean) and intraclass-correletion-coefficient (ICC).

Are there any meaningful further analysis that I can/should do to better understand the "difference" between the two devices? Would an ANOVA or Friedmann test be helpful? Or which test would be ideal in such a situation to compare the devices?

Thanks so much for your help.

Thanks so much in advance
 
Last edited:

Karabiner

TS Contributor
#2
Are there any meaningful further analysis that I can/should do to better understand the "difference" between the two devices?
You could maybe tell us what these devices are and what the measurement is, who the subjects in your sample are,
and how large your sample size.

I do not quite understand "I will have 3 measurements with device A, 3 measurements with device B (and eventually 3 measurements with device C), all of them at the same timepoint in X subjects.". So you perform 9 measurements at the same time?


With kind regards

Karabiner
 

fed2

Active Member
#3
i like to use lin concordance in this situation if i understand. 'statistical tools for measuring agreement' is the book. it gives models for many raters/measures etc.
 
#4
yes, there will be 9 measurements directly after each other (3 per device) in about 150 subjects. order of device will be random, so that not always the same device is used as the first one.
 
#5
I am not sure if lin concordance is correct, as none of the three devices is a gold standard, and as I am using three and not two devices. anova/friedmann would be correct, or any other idea?
 

Karabiner

TS Contributor
#6
better understand the "difference" between the two devices? is a a bit diffuse as objective,
but maybe what you are looking for is some intraclass coefficient (ICC).

With kind regards

Karabiner
 

fed2

Active Member
#7
weird fact that lin-correlation computed same way if gold-standard or not. standard error is different though.