12
$\begingroup$

This post about making books was posted to meta.SE very recently.

Of particular interest to me, though, was Monica Cellio's answer -- Mi Yodeya have made four books already, including one that had a print run. The total length of their four books comes to about 150 pages.

Notably, the use of content is pretty straightforward under the license terms (as long as any such publication follows the rules of course).

Making a book or books has been raised once or twice I think (not necessarily in a formal question here; but maybe in chat).

Is there some interest in perhaps trying to organize something like that, or possibly several such books?

Markdown seems to be relatively easy to turn into other forms now, I think the biggest challenges would be organization of content/curation/editing type tasks. [This might be tricky time-wise for some of us.]

I wonder if perhaps a more focused book rather than a collection of best posts might be a place to start -- e.g. there are some great relatively elementary posts that might be suitable as a book for people at an introductory level.

| |
$\endgroup$
  • 2
    $\begingroup$ I think one must consider the subject and the audience of the book. Mi Yodeva is truly revolutionary thing in many aspects. I am not Jewish and sometimes I browse it out of encyclopedic curiosity as I would otherwise never know this stuff. Also their audience does not often have access to this information. Our main niche and audience are... what..? With the exception of some expert opinions (yours included) we reproduce standard undergraduate Statistics courses to a great extent. Similarly our audience has access to a wealth of free authoritative resources (eg. ESEL). $\endgroup$ – usεr11852 Feb 2 '16 at 7:13
  • $\begingroup$ (I do not aim to come across nihilistic; I am just stating what I see as an obvious issue in undertaking such a task.) $\endgroup$ – usεr11852 Feb 2 '16 at 7:19
  • $\begingroup$ In fact I think just above elementary level is where a book would have the most value. (I don't see your comment as nihilistic; in any case you should feel free to express your thoughts) $\endgroup$ – Glen_b Feb 2 '16 at 8:01
  • 2
    $\begingroup$ That sounds very interesting. Some/quite a few of our answers are fairly directly adapted from other sources/textbooks. Would a book, rather than the posts here, be more likely to be seen as selling others' work under a different title? $\endgroup$ – Christoph Hanck Feb 2 '16 at 10:38
  • 2
    $\begingroup$ @Christoph There would need to be the usual care over copyright in the process of editing, but many works offer occasional short quotes of other works; note that anything that would contravene copyright law if we put it in a book already does so by being on site - and should be removed. (However, just by way of clarification -- we wouldn't be selling anything -- at least that's not my intent; making a book doesn't mean selling a book; under the terms of the site, people must be allowed to copy it freely and even reuse it for their own works if they follow the rules) $\endgroup$ – Glen_b Feb 2 '16 at 11:14
  • 7
    $\begingroup$ I think the real value of CV, at least to me, is that it collects together some of the subtle-but-important results from statistics/ML that are often just beyond the coverage of introductory material. I'm most familiar with my own posts, so please don't take this as self-promotion, but my answers on disadvantages to ROC AUC and why RF is not well-suited for GLM feature selection stand out as posts that would benefit folks with undergrad backgrounds in statistics -- important results that wouldn't make the cut for a first course in the topic. $\endgroup$ – Sycorax Feb 2 '16 at 15:44
  • 5
    $\begingroup$ @user777 I point people to your random forest feature selection answer all the time at work. $\endgroup$ – Matthew Drury Feb 3 '16 at 5:39
6
$\begingroup$

Glen_b, I like this idea a lot. Wedged between the ever-rising tide of questions asking about how to use R, there's a bounty of great statistical content on CV.

I don't know how much of that content is well-suited for introductory texts, though. There are a plethora of introductory probability and statistics textbooks (with a high variance in quality). But to my knowledge, there are not a great number of textbooks directly addressing how to do good predictive modeling with machine learning methods. The textbooks which do exist tend to fall along the lines of "a first course in predictive modeling." These fill a niche and are certainly necessary, but after completion, there are not great subsequent resources. It's usually in this space where the difficulty of the material suddenly jumps like it found a spider under the toilet seat. People get stuck in the local minimum of looking at models with some particular flavor (trees, Bayes, kernels, neural nets), but being a good practitioner means knowing how to use appropriate tools for the problem at hand, and that point seems to get swept under the rug of the oft-mentioned-but-rarely-examined "domain knowledge."

I suppose I'm imagining a book with the subtitle "What is good decision-support, and how do we accomplish it?"

I think the best form a book could take on would be a tour of the "black arts" of predictive classification, i.e. the topics that are incredibly helpful to practitioners but are rarely discussed in a single location.

  • Presently I'm putting together a presentation on the use and abuse of different performance metrics. I was surprised to find that the material I'm drawing from is scattered across a diversity of snippets from different articles, but not any single resource. And my interactions with Prof. Harrell underscore just how subtle the question is.
  • My highly-upvoted post on Euclidean distance high-dimensional spaces is actually borrowed from an article on subtle-but-important points in the machine learning tradition. IIRC, that article was published 15 years ago, but its points remain just as obscure today as they were then... at least, in my anecdotal experience.
  • If I had an up-vote for every time I spoke to an engineer who wanted to use random forest to do feature selection for a linear model... well, actually, I just might.
  • What are the use cases for uncalibrated models? When is using ROC ACU obviously and provably the wrong choice?
  • "Why does everyone at the office get snippy when I mention gradient descent?"
  • Machine learning tools tend to work best with atomic units. What are the options for working with data that is non-atomic? Networked, geospatial, time-series, clustered, hierarchical?
  • This is probably neither here nor there, but nothing I ever learned in a linear algebra classroom has ever helped me actually do linear algebra in practice. Knowing how to set up your pipeline of functions and analysis so that you have convenient factorizations stored for future use in computing inverses is huge!
  • Hyperparameter optimization: most people use grid searches, but there are more intelligent ways to go about this, such as using surrogate models.

But these are just my observations from my very narrow view into modern applied statistics. My perspective, like anyone else's, is constrained by context, so perhaps I'm just making a big deal out of things that are important to my role.

I guess what I'm getting at is there's considerable value to the scattered uncommon insights that the specialists at CV have touched on.

Also, I might just be cranky.

| |
$\endgroup$
  • 5
    $\begingroup$ At a purely selfish level, I have little doubt I would benefit from such a document. $\endgroup$ – Glen_b Feb 3 '16 at 21:41
  • 1
    $\begingroup$ @Glen_b Yeah, I would, too. I mean, maybe such a book doesn't in the same way that mathematics can prove the nonexistence of certain objects. But I'm contrarian and enjoy charging at windmills, so... $\endgroup$ – Sycorax Feb 3 '16 at 21:55
  • 2
    $\begingroup$ I do like your idea, but also think that we've plenty of stuff at an introductory level that would complement textbooks nicely. In particular: posts addressing common mistakes & misconceptions (which may be passed over, or even promulgated, by textbooks); & those tidbits that broaden rather than deepen knowledge, e.g. How to find a good fit for semi-sinusoidal model in R?. $\endgroup$ – Scortchi - Reinstate Monica Feb 4 '16 at 12:30
  • 3
    $\begingroup$ @Scortchi That's true. I think that my biases are showing here -- I tend to work on the kinds of problems discussed in the post, so I'm inclined to think they're important. At the same time, there's no reason we have to stop at 1 book! $\endgroup$ – Sycorax Feb 4 '16 at 13:35

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .