A maximally simplified presentation of how metrics are random variables, and how they can be gamed. Uncorrelated variables will produce a correlation in samples.
A maximally intuitive presentation on what correlation is not, with maximally simplified concepts. [Note that I improvise 100% and I don’t prepare] There is a more technical lecture to come.
The technical paper is here: Fooled by Correlation: Common Misinterpretations in Social “Science”
SIMPLIFIED TUTORIAL, 1 (in the series where we break down concepts intuitively):
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!
See the whole book (gets technical beyond Chapter 5)
A VERY SIMPLIFIED TUTORIAL (VERY SHORT)
Correlation measures are misused in the presence of nonlinearities.
(How a measure of unintelligence can give the illusion of high correlation with performance.)
Correlation measures are not supposed to be used in the presence of nonlinearities. When 2 variables correlate half the time (in a symmetric way around the mean), correlation will not be 50% but will show ~90%. Part of debunking IQ studies. If IQ works for disabilities but does not correlate with success, there is an illusion of correlation because of the biases in the metric.
How the correlation of subsections will be lower than the total (in absolute value). With implications for “empirical” research.
Link to PDF – https://www.dropbox.com/s…
Free copy of the paper – https://arxiv.org/pdf/1802.05495.pdf