Graphical explanation of the important and common statistical paradox called Simpson's paradox ... https://www.youtube.com/watch?v=39RZFm4EEzQ

In a randomised controlled trial to determine if treatment T reduces mortality for people with disease D: 3 out of 241 in the treatment arm die (1.24%) 10 out of 249 in the control arm die (4.02%%) So is the treatment effective in reducing mortality? Classical statistical p-testing says no and led to the conclusion "The mortality rate in the treatment group and control group are similar". Is this reasonable? The Bayesian alternative suggests not ..... ... https://www.youtube.com/watch?v=QX8Nq4fr9g8

For those who do not know Simpson’s paradox and why it is so important in understanding and interpreting many observational studies, this 10-minute video explains the paradox using very simple hypothetical data based on the recent Canadian study that showed “Unvaccinated more likely to be injured in a car accident than vaccinated”. The authors of that study hypothesized that this was due to inherent mental and social problems of the unvaccinated. But they did not consider the simple explanation that the unvaccinated were more likely to be driving during the lockdown period since they were banned from planes and trains. My previous video showing why the Canadian study was stupid: https://youtu.be/_iryCrHaozU The Canadian study: https://doi.org/10.1016/j.amjmed.2022.11.002 My previous videos on Simpson’s paradox: https://youtu.be/39RZFm4EEzQ https://youtu.be/2Dz6XPjD7YE ... https://www.youtube.com/watch?v=hKLYrg--0Sg

A short (5 minute) introduction to why purely data driven machine learning algorithms are cannot properly handle most real-world decision problems ... https://www.youtube.com/watch?v=-7vSiWRasxY