Statistical correlation is not sufficient for causal reasoning without a structural model.

“The flaws come if you try to impose causal logic on correlation. It doesn't work too well.”

Data interpretations, such as average class sizes, can be corrected by identifying and reversing sampling bias processes.

“If you know what that sampling process was and you can model the bias. Sometimes you can invert it. Sometimes you can reverse that filter and make an an unbiased estimate from biased data.”

Bob believes that his wife knows a lot about distribution curves.

“no the graph it's it's it's a distribution curve Bob's wife knows a lot about them”

Wade suggests that the smart and dumb sides of a distribution curve are essentially the same.

“you always pick the dumb side because they're going to be right cuz it's the same it's the same as the smart side in in the graph”

Frequentist statistical tests like P-values and confidence intervals are often meaningless and misused because they do not provide the probabilistic answers that people assume they do.

“it's all based on this yeah these these ridiculously I'm not going to say flawed they're just they're just a meaningless way of doing of doing analysis of Uncertain data especially if they're used to say that something works right like that's the ultimate question”