- Plot errors against y-hats
- What is outlier test and KS test?
- Use function predict() on GLMM object. Predicts bitter values based on model parameters. Predictive posterior
    - predict.glmmTMB()
-ggeffects package
- vizreg package
- Ensure errors look normally distributed
- Check that glmmTMB() spits out an error if it doesn't converge
- Use hatvalues(model) to look for leverage points
- Do predictions for each separately, see if any are fundamentally not giving me aroudn the right number of bitter/non-bitter
- Also compare weighted models (top 4 or top 8) to single models! Factor in number of parameters (less is better)