2
$\begingroup$

So I've been wishing to learn about $p$ value and hypothesis testing for a while now. But since my application is mainly in the domain of ML, and these concepts virtually never shows up in any of the textbooks/applications, therefore I am having a hard time finding the motivation to learn these concepts.

For instance, you never hear "let's do hypothesis testing with convolutional neural networks..."

Then this got me thinking, what are some commonly taught concepts in statistics which are not found in ML literature/publication/discussion?

And why is that the case?

Since I don't have a background in statistics, therefore can someone familiar with this issue please chime in and shed a light?

Improve this question
$\endgroup$
4
  • 6
    $\begingroup$ I am curious to see responses but also fear that this is too opinion-based to be a good fit for Cross Validated. After all, there is far from a consensus about where machine learning ends and statistics begins. People also have very different approaches to learning machine learning, and it would not be so unusual for someone to jump straight into Keras without learning OLS linear regression, yet plenty of machine learning practitioners know OLS just fine. $\endgroup$
    – Dave
    Commented Jul 8, 2023 at 1:10
  • 1
    $\begingroup$ This question does not require or develop statistical expertise in the way that I would expect for a good question on this site. But I asked ChatGPT: “What are some common concepts in statistics that are not used much in machine learning?” Since I got a reasonable answer, I recommend that as a way to start. Have you asked there or somewhere similar, and if so what did you think of the response? $\endgroup$
    – user225256
    Commented Jul 8, 2023 at 1:24
  • 1
    $\begingroup$ You might find this other question and its answer interesting, relative to why oversampling/weighting as in survey design is not used in machine learning: stats.stackexchange.com/questions/611423/… $\endgroup$
    – J-J-J
    Commented Jul 8, 2023 at 6:47
  • $\begingroup$ In addition, it might be worth asking your question on hsm.stackexchange.com , though I'm not familiar with this other website. Anyway, there may be historical/sociological causes to some statistical concepts not being taught in ML textbooks. While ML and statistics practitionners & researchers may have an essential insight into this, a social scientist could investigate it more systematically. $\endgroup$
    – J-J-J
    Commented Jul 13, 2023 at 6:04

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.