Bayesian Vs. Frequentist
Statistics: Are you Bayesian or Frequentist? is a great introduction to some foundational understanding behind the concept of probability, about which I remember I first read in How to Measure Anything, which is obviously a must-read.
Most of us take probability for granted, but the devil is always in the details. And in this case, the devil is rather philosophical.
The video is fascinating, and the following explanation thought-provoking.
The Scientific Method
Science in a High-Dimensional World is an enticing rumination on the scientific method:
In a high-dimensional world like ours, there are billions of variables which could influence an outcome. The great challenge is to figure out which variables are directly relevant – i.e. which variables mediate the influence of everything else.
A remarkable empirical finding across many scientific fields, at many different scales and levels of abstraction, is that a small set of control variables usually suffices. Most of the universe is not directly relevant to most outcomes most of the time.
With that understanding:
In general, the goal of science in a high dimensional world is to find sets of variables which mediate the influence of all other variables on some outcome.
Overcoming the Pervasive Analytical Blunder of Strategists provides an important point of view:
There Can Be No Data About the Future
Modern business education trains all students to believe that good decisions are based on rigorous analysis of data…
A critical tenet of the inferential methodology is that if a conclusion is to be drawn about a given universe, the sample analyzed must be representative of that universe.
So, a good decision is made on the basis of inference from the rigorous analysis of data drawn from a sample that is representative of the universe with respect to which we seek to make decisions.
The problem is:
What is generally not taught explicitly is that the inherent data limitation is that 100% of the world’s data is from one era: the past.
So sometimes it works:
If we can say that the pool of data from the past is indistinguishable from the pool of data from the future, then we can be confident that the data from the past is completely representative not only of the past but the future as well.
But sometimes not when:
…a sample drawn from the past is utterly unrepresentative of the universe including the future. And it is particularly problematic in being unrepresentative in ways we cannot know until after the fact.
The risk is high:
Despite Aristotle’s very specific warning to never use analytical inference in the part of the world where things can be other than they are, and Wolpert’s modern restatement of the warning, not only does the business world routinely violate the principles of statistics and sense to do exactly that, but the learned institutions of business education also teach the error-filled practice with enthusiasm bordering on fanaticism. It flat out disables otherwise intelligent men and women. It deludes them into thinking that their choices are valid when they are, in fact, designed to be fallacious blunders.
A good advice:
As a strategist, before you make any decision based on data analysis of any sort, you must ask yourself one simple question: Am I willing to assume that the future will be identical to the past? If the answer is yes, then do what the analysis tells you to do. If the answer is no, then DO NOT USE THE ANALYSIS.
Explaining p-values with puppies explains:
A p-value doesn’t prove anything. It’s simply a way to use surprise as a basis for making a reasonable decision.
A p-value asks, “If I’m living in a world where I should be taking my default action, how unsurprising is my evidence?” The higher the p-value, the less ridiculous I’ll feel about persisting with my planned action. If the p-value is low enough, I’ll change my mind and do something else.
Sleep and Death
How is depression like sleep, and why does sleep deprivation treat depression?
Melancholic depressive patients report that they feel worst in the morning, just after waking up, get better as the day goes on, and feel least affected in the evening just before bed. Continue the trend, and you might wonder how depressed people would feel after spending 24 or 36 or 48 hours awake. Some scientists made them stay awake to check, and the answer is: they feel great! About 70% of cases of treatment-resistant depression go away completely if the patient stays awake long enough. This would be a great depression cure, except that the depression comes back as soon as they go to sleep. There’s a lot of great work going on to figure out how to make cure-by-sleep-deprivation last longer…
The whole article is a real treat of thinking.
Goofus and Gallant
A Failure, But Not Of Prediction points out something critical:
Making decisions is about more than just having certain beliefs. It’s also about how you act on them.
It’s an old article dated April 2020, but I think it’s worth reading anytime.
…a Short Treatise on the Nature of Failure; How Failure is Evaluated; How Failure is Attributed to Proximate Cause; and the Resulting New Understanding of Patient Safety
Techniques of Science Denial
A denialism taxonomy introduces FLICC:
- Fake experts
- Logical fallacies
- Impossible expectations
- Cherry picking
- Conspiracy theories
The denial industry has a well-developed and constantly evolving playbook. Wealthy interest seeking to sow doubt about reality — about /whether reality can even be known/ — can pay for skilled denialists to plan and execute denial on their behalf.