Hi everyone,

I have a question relative to an ICC analysis. I want to compare 2 methods to assess motor blocks occurring during walking in Parkinson patients. The first method is the clinical eye from an examiner and the second one is an automatic detection of the blocks based on an accelerometer sensor. We tested 19 patients while they were performing 5 different walking trials in 4 different conditions: with and without medication and with and without a walking aid. So we have a total of 20 trials. Our outcome measure is the percent time spent blocked over the total time spent walking for each trial. And I want to compare using ICC the automatic detection method, to the gold standard method consisting in the clinical eye of an examiner to check whether the automatic method is reliable. But I'm not quite sure whether I should compare the 2 methods per trial (19 patients * 20 trials = 380 cases), per condition (19 patients * 4 conditions = 76 cases) or per patient?
This is crucial especially because it seems that these different methods give different ICC results.

I'm looking forward to reading your opinion on that.