Whoaaa you're amazing! Yes 'steps' contains 5 zeros! That it should't have obviously but allright! Thanks so much i am gonna try to delete the 0 rows and then see what happens!
Exciteeeed this might work!
These are a lot of exclamation marks but you'll hopefully notice my gratitude.
As for the other part, i don't know how to put two the same independent variables in the model and then see which one works. However, i was thinking of making two models for log(ketosis) and steps. One being log(ketosis) ~ steps (-3, -2 and -1), and the other being log(ketosis) ~ log(steps) (-3 -2 and -1). And then using AIC to check which one is better?
Does that make as much sense as i think it does or is this the misinterpretation of someone who is educated for figuring out what is wrong with animals instead of data?
Then: if i can use the AIC scores, can i just do one comparison (one with log(steps) and one with normal steps) and say: hey, AIC score for (for example) normal steps is lower than the model with log(steps), so based on this i can say, for all the models (i have different ones with different random factors and different outcome variables) i can use normal steps? (if it wasn't obvious by now, english is not my native language and i am struggling to explain myself properly).
Thanks!