StatQuest: Bootstraping
During the "Numerical Methods I" course together with students at Faculty of Physics we talk about experimental data fitting. Matlab's polyfit function will cover most common use cases, but in Physics we are often interested not only in the point estimates of measurements. We also care about associated errors and the documentation of polyfit function fails to deliver on that. Unless you are well versed and statistics and know that you need to calculate covariance matrix and take square root of values on its diagonal.
I was caught of guard by the students who were interested in actually estimating the measurement error using polyfit. As at the time I wasn't prepared, I have told them about Bootstrap method, which I like a lot. I like this method, because it has natural interpretation (it gives us a confidence interval) and is applicable even in highly complex situations, for which no analytical error estimation formulas exist.
More about this method in the following video by StatQuest with Josh Starmer.