Before we talk about correlation, let's discuss standard deviation, its analog in dimension 1. People don't get while using it as a metric for deviation!
What are Fat Tails? This is very introductory.
Everything in empirical science is based on the law of large numbers. Remember that it fails under fat tails.
The CLT allows anyone (including ignorant economists and psychologists) to do statistics by using prepackaged recipes coming from the Gaussian. What are the foundations? How does it work? Where does it not work?
A maximally intuitive presentation on what correlation is not, with maximally simplified concepts.
A maximally simplified presentation of how metrics are random variables, and how they can be gamed. Uncorrelated variables will produce a correlation in samples.
We saw that 1) many metrics are stochastic, 2) what is stochastic can be hacked. This is the simplification of my work showing that "p-values are not p-values", i.e. highly sample dependent, with a skewed distribution. For instance for a "true" P value of .11, 53% of observations will show less than .05. This allows for hacking: in a few trials a researcher can get a fake p-value of .01.
Power laws, extremely simplified.
Q&A on MIN-LESSON 8 (Power Laws). Why violence did not drop (the Pinker Problem) -- Pandemics --What an infinite mean means -- What causes power laws.
1) The law of large numbers (properties of aggregates) works in one direction. Why you can generalize from particulars, never particularize from generals. 2) The difference betweern clinical, statistical, and risk management approached. Why they don't scale. 3) Never compare Mediocristan to Extremistan (Covid to car accidents).
Why the statistical properties of a group will NOT particularize to any individual member of the group. Explained in 2 minutes. The example uses central limit.
A simple case study where model error shows the claims in a paper published in a "prestigious" psychology journal to be false. Why worry about low R^2.
In every age bracket, the vaccinated live longer than the unvaccinated (using all cause mortality). However as a group the unvaccinated appear to have a longer life expectancy. This is because the vaccinated tend to be older (hence more likely to die). I explain Simpson's Paradox in general.
Explaining path dependence and maximum drawdown.
Quick presentation of drawdowns and the necessity to use logarithms for returns.
How to look at the risks of Covid vaccines, why they much lower than you think. We never had a larger monitored sample size in history and it allows events that on average show up later to manifest themselves very early on.
Violence is from Extremistan, hence requires some more sophisticated tools since LLN works slowly. We see how Pinker's thesis is bogus. We look at ways to integrate the factual unreliability of historical accounts. We look at transformations to analyze violence using power law tools since the worst case is bounded at contemporary population level.
What do you do when the data looks like it is powerlaw distributed over a broad range, but cannot be technically a power law? We use a dual distribution and transport parameters between one and another.
A first, very introductory presentation of fragility as linked to both nonlinearity and dislike of variations. Antifragility is almost the opposite, limited to a specific range of variations.