Sunday, March 25, 2018

How might I obfuscate, let me count the ways

A reasonably clear article on the normal mainstream media statistical innumeracy. In this instance, it is about gun ownership and murder but the same pattern of errors shows up all the time in innumerable other reported debates. From Everybody's Lying about the Link Between Gun Ownership and Homicide by BJ Campbell.

The usual sins committed by journalism majors (and ideologues) are:
Cherry picking data
Omitting outliers
Omitting R^2 Value
Omitting effect size
Committing the category error of mixing data populations (ex. including suicides with murders)
Changing data definitions mid argument without noting the change.
Omitting context
Use of proxies when direct measurement is feasible
Confusing correlation with causation
Obscuring ideologically uncomfortable causal information
Omitting context
Omitting alternate explanations
Not illustrated here in this article but otherwise common errors include
Anecdote over data
Too small sample sizes
Non-random samples
Reliance on self-reports over observed data
The world is hard enough to understand as it is without all the orthodox cognitive pollution which obscures reality.

No comments:

Post a Comment