Without wanting to look at the entire data and coming to my own conclusion, is the sample size big enough to make any estimations on the effects of the trainer level?
I've been working on this and age, and both are getting there, but are still pretty noisy. The age one should be good enough to release pretty soon, but trainer level is rough.
Specifically, we have a ton of data with level 4 and 5 trainers (there's 22 training type-skill combinations that train at a high enough rate and have enough data to compare the training rate between them), but level 6 is still very iffy. Levels 1-3 and 7 are basically non-existent (1-2 users with level 7, none with levels 1-3). So the only comparison I have any confidence in for now is the level 4 vs level 5 one. So all of you hoarding those level 6 and 7 trainers should join up