Great video Emma! Very digestible and clearly explained. Some additional context here for those who are interested; At @8:56, the bootstrap does rely on some assumptions. Ultimately, the consequence of these assumptions is that the standard Bootstrap struggles with extrema statistics like the sample maximum and sample minimum. So, you'll want to keep that in mind during interviews. Also, at risk of adding additional confusion, the Bootstrap is generally considered a non-parametric procedure. Alternatively, there is a parametric Bootstrap. Either of these can be used to construct confidence intervals whose interpretations can be equated to hypothesis tests, and in practice often are.
Wow, thank you very much for the kind compliment! It makes me so happy to know you're benefitting from my content, which is the whole reason I make it in the first place! Thanks for watching! 💛
Hi, I recently came across a question. What is the probability of getting all unique samples when bootstrap sampled n times in dataset containing n datapoints.
Imagine you have an unknown population of positive real vectors (ai, bi, ci, di). You are given a sample from that population. You don't know what sampling algorithm was used, but you know that it was a biased algorithm in that vectors with small first component 'ai' were more likely to be sampled. Based on this biased sample, is there a way to a) make an educated guess at what the sampling algorithm was, and b) find a sampling algorithm that correlates even more with small first components?
Hi Yash, thanks for your comment! We don’t currently release our video content in the form of slides. However, if you’d like to know about some of the other services we *do* offer, feel free to check out our website at datainterviewpro.com to learn more! 💛