Thanks for your reply! You are totally correct with your assumptions. I’ve discussed the issues of different ratios and multicollinearity a bit in the other post I linked above. It seems the opinion of including or excluding the intercept or the third ratio diverge. I guess I can easily test the effect of both by doing model comparison.
Anyways, the most pressing issue for me at this point is how to do model the effect of total area on the uncertainty of the outcome. If your think bf(diseased | trials(1) ~ type1_ratio + type2_ratio, phi ~ total_area) is reasonable, I’ll continue with checking this route.
Thanks again