statistics is the most beautiful and egalitarian discipline of all of those i have studied, because when we are confused about anything in statistics we are almost always confused about the elementary, foundational concepts

in my personal experience this is not true when learning languages, history, economics, or other branches of mathematics

Example 1: sometimes very smart people suggest nonlinear regression as an alternative to quantile regression, when these address COMPLEYELY different problems. the former relaxes the linearity of the conditional mean function and the latter relaxes the focus on the mean itself.

Example 2: the smartest people on econ are usually deeply confused about quantile regression and mostly fail to understand what a conditional quantile is. this is because we do not actually understand what a cumulative distribution function is.

Example 3: we freak out if covariates don't end up well balanced in our experimental samples because we don't understand what an expectation is. we think it delivers more than it really does.

the best thing about teaching econometrics is every semester some students come to my office hours and say "wait, but if THAT'S what an expectation is then why would i ever care about it?" and i get to be like "congratulatios. you are now asking the correct questions."

example 4, bonus example just about me: I can't intuitively grasp splines and this is because i don't understand the design matrix well enough in the linear regression case

example 5: "But we aren't IN asymptopia!" the asymptotics are only EVER there to provide an approximation to finite sample behaviour because nobody is smart enough to get exact understanding of general finite sample behaviour

example 6: nobody understands weak IV asymptotics because nobody understand asymptotic analysis is always a simplified model designed to produce an approximation, and if the root N asymptotic approximation sucks, you can just change to some other power of N

(so it ought NEVER to comfort you that you can prove such and such asymptotic behaviour to be true. There are many, many true asymptotic results. The question is not whether the asymptotics are true or not but whether they are USEFUL, appropriate, good, etc.)

this thread brought to you by: me drawing the PDFs of a fair die and a loaded die, and then fucking it up the first time i try to translate my loaded die's PDF into a CDF

@KashRamli this is for u