It may be better to use as a two-sided log-normal distribution (i.e. a regular log-normal distribution combined with the same distribution flipped over the Y axis). Unclear about the details.
Break down effects on factory-farmed animals into multiple categories for different types of animals
Might be overcomplicated and not worth it
Write function to sum multiple distributions where you specify at runtime which are positive or negative
see http://www.nickbostrom.com/fut/evolution.html
For posterior, represent estimate as buckets instead of converting to lognormal distribution, which loses accuracy
No known bugs right now. Yay!