No I think you misunderstand. There's a data-generating process: people commit crimes. For a given set of people, the mix of crimes they commit will drive both

No I think you misunderstand. There's a data-generating process: people commit crimes. For a given set of people, the mix of crimes they commit will drive both convictions and killings. For instance, if the mix of crimes a particular group commits has more murder in it than for other groups, we would expect both more police-shootings of those people (because police can intervene with deadly force when lives are at stake) and more convictions for murder. You're right that people who are shot by police don't get convicted of anything, of course.