Engineering Statistics Handbook | Hacker News Reader

Engineering Statistics Handbook

itl.nist.gov

145 pointsbrudgers13 comments

I have to say, NIST dudes are/were so underappreciated by the likes of me. Is it just me?

I was in a LUG in the DC area and one of the dudes there was a NIST guy. His use of Linux and Unix was not surprising, but when he would tangentially mention what they were doing with their boxes I would always wonder: how the hell do more people not pay attention to NIST? But maybe that is just me. I have done a lot of sysadmin in the education industry, where we are famous for listening to no one and being the butt of security nightmare jokes.

As someone who is weak in all post-high-school maths, and stats in general, I have definitely bookmarked this for later. I did not care for stats until I finally took my first true computation linguistics class, Statistical Methods for NLP, and years of wannabe linguistics without math seemed pretty retarded after that.

Thanks for a cool link.

It's not just you.

I don't know how are things in the US, in Europe not many people even know they exist. 10 years ago I was doing some work with biometric identification and it turned out these guys had an open source system for fingerprint recognition (including test data).

Not only they were incredibly helpful, they even sent me a couple of CDs with all the information for free.

As good as that is, see this: http://dlmf.nist.gov/, the online companion to the truly epic NIST Handbook of Mathematical Functions, itself the modern successor to Abramowitz and Stegun.

That is a nice piece of work. If you click on any of the 3D graphs you can grab them and drag to view from different angles. Very cool.

It (and the book on which it is based) are amazing. Virtually every older applied mathematician has a copy of the original Abramowitz and Stegun, which was supported by the National Bureau of Standards, prior to NIST. The successor (http://www.cambridge.org/us/academic/subjects/mathematics/ab...) is really an incredible achievement. The reviews are hilarious. One notes "The book is quite heavy; for convenience, one might be inclined to place it on a stand, as with an unabridged dictionary". Another review begins "This is like trying to review the bible".

The online version also lets you get LaTeX for everything, and has many other features. But the print version is a thing of beauty.

In full disclosure, the Editor in Chief and Mathematics Editor, who devoted the last 13 years of his professional life to the project, was my stepfather.

I worked at NIST about 10 years ago, for a summer internship in the Physics division.

Absolutely top notch scientists, some good facilities, a decent budget and very well scoped research objectives. Management was also largely hands-off. Great place to work.

Looking at the chapter on Exploratory Data Analysis (EDA), it mentions that "EDA techniques are generally graphical. They include scatter plots, character plots, box plots, histograms, bihistograms, probability plots, residual plots, and mean plots."

There is another term for EDA, and it has a whole industry built around it: business intelligence (BI). BI is focused on aggregating data from operational systems (OLTP) and producing new systems optimized for analytics and decision making (OLAP). Top products in this space include: QlikView, Cognos, Tableau, Hana, etc. One can make a decent hourly rate working as a consultant in these systems.

Okay, I read through the first few pages and it seems to me that this is a difficult presentation, where facts are presented before they are explained.

Handbooks are typically constructed for quick reference by experienced practitioners of the subject. Handbooks favor breadth over depth and this means they tend to be more enumerative than explanatory.

This is a common difference between presentation for reference and for pedagogy.

It is extremely difficult to write well for both purposes, so good examples are rare (c.f. Kernigan & Ritchie)

I agree about the difficulty in serving two masters and that The C Programming Language is well written in this regard. Most authors are unfortunately not in position to authoritatively declare portions of a subject undefined and get away with it. K&R is not really so much a handbook as a specification. Its strength is that it often specifies by showing rather than describing.

This brings back memories of when I would program in R at undergrad. Bookmarked.

Of course most of the so-called engineers who read this site (programmers w/out engineering degrees) have no use for stats.

Reply on news.ycombinator.com