Krippendorffs’ alpha - sample size

#1
Hi all, I’m not good at this stuff, so please bear with me.

I am doing an investigation where I want to look at agreement between different groups of raters.

The raters (a group of novice/new raters) and a group of exert raters will watch participants before an action and grade that (probably) on an ordinal scale of 1-3(ish).

Similar research uses kappa for the individual scores, but they normally have very low raters number (i.e. 2!). I believe that Kappa could be influenced if multiple users are used?

Another paper that used multiple raters used Krippendorffs’ alpha, so, that’s why I’m thinking of using that. They then use ICC for the total scores.

However, I can’t find anything re: sample size estimation for Krippendorffs’ alpha.

Also, and I’ve not been able to find the answer anywhere of this, but what is my “sample”?
  1. is it the number of raters (i.e. 20 if x10 novice and x10 expert)?
  2. Is it rater groups (so n = 2 no matter how many are in each group)?
  3. Is it based on data points, the number of tests scored? I.e. 2 raters watching 100 different participants will produce more data than 10 raters watching 5 participants