Originally Posted by I-Like-To-Bike
When the accident data gives no indication of the injury severities (or if there even was an injury) and just totals up the number of collisions, falls, and boo-boos as if they were identical I am saying any conclusions drawn about relative danger/risk are of NO value. And in fact, are reason to suspect either the intelligence or honesty of the person who attempts to make such conclusions.
I decided against using the EDIT function ...
More generally, I think that using more suspect data is tolerated for questions where the data is limited. If we had access to "golden data" with little error and biases, then I would be more likely to agree with you. However, we are discussing an issue where many strong opinions are formed by our individual anecdotal experiences and the data is quite limited. My suggestion is that we use what is out there to make some inferences about reality but then be "real" about the strength of our evidence.
In this sense, I think Forester fails at times to put his results in the proper context or overstate the results.