stock here, submitted by ACD
I discovered the new Modernism at a family birthday party. The deniers of reality who want to suck down every Government and Media narrative, are now reduced to:
- There is so much conflicting information, how can we really know ANYTHING
- Nothing is either true or false, everything is a spectrum
- There is no right or wrong, and everyone is entitled to their opinion. Correction, there is no wrong, unless you disagree with me, because my feelings and virtue are so clear
Epitaph for this empiricist: “There are no accidents in the Universe,” replacing the old saw about GOD playing dice.
Coinky-dinks, however, abound. For example:
An investigation of official statistics has found that the number of athletes who have died since the beginning of 2021 has risen exponentially compared to the yearly number of deaths of athletes officially recorded between 1966 and 2004.
So much so that the monthly average number of deaths between January 2021 and April 2022 is 1,700% higher than the monthly average between 1966 and 2004, and the current trend for 2022 so far shows this could increase to 4,120% if the increased number of deaths continues, with the number of deaths in March 2022 alone 3 times higher than the previous annual average.
For those of you still bothered by — or harassed by — the hoary notion “Correlation does not necessarily imply causation,”, consider the seminal work of Rosenbaun and Rubin, “The central role of the propensity score in observational studies for causal effects,” Biometrika. 70 (1983) 41-55.
The flip side of the statistical coin: “Causation necessarily implies correlation.” As I did in the arena of “motor vehicle risk analysis”, wherein “accident causation” proved NOT an oxymoron, one can apply Propensity Score Analysis to good advantage with existing datasets underpinning studies “criticized” or “labelled” CORRELATIONAL. Transforming a pig’s ear to a silk purse has never seemed simpler, in my opinion! If this comment board supports even more verbiage than the preceding, I provide an abstract of a study (2004) I presented to the Society for Risk Analysis.
Motor vehicle collisions are complex events involving multiple factors related to drivers, driving environments, and vehicles, a truism often belied by statistical estimations of risks of real-world crashes and their consequences. Many published studies feature logistic regression analyses of police-reported (observational) data on crashes and appear prone to biases due to omitted relevant variables and failure to deal effectively with confounding. In most cases, however, analysts ask seemingly simple, cause-and-effect questions. For instance, did this change in design, or would this change in condition, mitigate the target problem? The goal is neither to model the complexity of “the problem”, nor to assess the importance of a certain risk factor relative to other factors. Other risk factors, however influential or relevant, can be considered nuisance factors. In such cases, propensity score analysis (PSA) may well be the method of choice. The purpose of this paper is to foster a greater appreciation of the utility of PSA for motor vehicle risk analysis. Two applications of PSA are (1) assessing the effect of occupant loading on the risk of rollover of 15-passenger vans; and (2) evaluating the effect of retroreflective tape on heavy trailers for the risk of side or rear impact by other vehicles during hours of darkness. Results of PSA indicated lesser effects of putative causes (higher numbers of passengers, presence of retroreflective tape) than claimed in published studies to date. In both instances, differences in the magnitude of effects were due mostly to the influence of omitted and confounding variables. Applying PSA, analysts still have to do thorough exploratory data analyses. For the preliminary assessment of simple cause-and-effect relationships, however, analysts can concern themselves much less about the structure and complexity of motor vehicle crashes and their outcomes.
————————————– Amazing Polly on Doctor Deaths