I am looking into race and homicides in the USA, 1995-2016. I want to look into "trans-race" (victim and perpetrator different races) homicides by race and year. However I want to also include effects of different racial frequencies within the population. Since it's count data, I was presuming to us a poisson family.
It's almost as if I am looking at a "double offset", where I offset the (log) count of "trans-race" murders by (log) total murder victims of that race, but how do I account for differences in overall frequencies of each racer?