Originally Posted by genec
In order to determine "likelihood," you need to determine the aggregate… This is typically done in the insurance business and is known as “actuary,” or the analysis, evaluation, and management of statistical information.
Well that's pretty vague, but, in general, I agree. Of course, the aggregate you need to determine depends on what likelihood you're looking for.
In particular, if you're looking for the likelihood of death from a given type of hit, then the aggregate data you need is all hits of that type (fatal and non-fatal), and to know how many of those hits resulted in death. As I've pointed out for too many days and too many posts now, I think you're confusing this, determing the likelihood of death given a hit, with what we need to determine to evaluate whether a cyclist is more likely to die from hit-from-behind or hit-from-elsewhere.
If you're just comparing the likelihood of
hit and death from one type of hit to the likelihood of
hit and death from another type of hit, which is what we're doing when we're trying to determine which type is more likely to kill a cyclist, then the non-fatal data from either type is completely irrelevant. The only aggregate you need is the total number of deaths from each.
Originally Posted by genec
All your current data says is that more people do die in collisions from “elsewhere,” but it does not determine “whether being hit from elsewhere is much more likely to KILL you [than is getting hit from behind in daylight].”
It most certainly does determine that. The way you determine relative likelihoods of two events is by comparing the incidence of those events. In the case of "killed from being hit from behind" the incidence is much lower than that for "killed from being hit from elsewhere", therefore being hit from elsewhere is much more likely to kill you than is being hit from behind. How much simpler can it be?
But even if the available data didn't determine which was more likely to kill you, how would knowing "how many people have been hit from 'elsewhere' and simply walked away", which has nothing to do with deaths, provide the necessary data for determining “whether being hit from elsewhere is much more likely to KILL you" than is getting hit from behind?