Print Story [+] [P] Books that give hands-on examples of how to use statistics properly
By nlscb (Thu Jul 15, 2010 at 11:38:30 AM EST) (all tags)
Are there any books out there that provide hands on examples on how to use statistics properly? I am not talking about how to the various proofs and exercises that under lie basic probability theory and statistics. I have already taken several calculus based courses in that. I am interested in ones that show how real life data was properly and correctly analyzed. I am trying to find a way to gain comfort with what I have studied through repeated practice.

I think an analogy to this would be the difference between studying physics in a text book and actually doing a lab experiment in physics. You can read all you want about physics and learn the theories, equations, and proofs. However, you are probably never going to get much comfort with it until you apply it to something in the real world. When you a do a lab experiment, thousands have done the same experiment before hand. You are not discovering anything new. You are just getting comfortable with using the scientific method to confirm things.
< A Day in the Life -- Patchwork | I saw that one coming >
[+] [P] Books that give hands-on examples of how to use statistics properly | 6 comments (6 topical, 0 hidden)
Since you also mention physics... by ana (2.00 / 0) #1 Thu Jul 15, 2010 at 11:54:50 AM EST
The old standby for years has been Bevington (Now with additional authorship of Robinson!).

It depends, of course on what kinds of statistical problems you have; the ones in social science tend to be different from the ones in the physical sciences, for example.

"And this ... is a piece of Synergy." --Kellnerin

How much math does that book require? by nlscb (2.00 / 0) #2 Thu Jul 15, 2010 at 12:14:04 PM EST
I have taken up to and including introduction to real analysis and introduction to partial differential equations.  I just barely passed those.  If that's not enough, what would I have to study to understand that book?

Also, how clear is it?  I've had a lot of trouble with math/stats textbooks that say things in the first introduction to new topics "If given premise A, one can quickly and easily see that conclusion Z flows from it" while leaving out steps B to Y.

Regardless, I will definitely take a closer look.  Thanks.

[ Parent ]
Not all that much. by ana (2.00 / 0) #6 Thu Jul 15, 2010 at 01:05:11 PM EST
Some integral and differential calculus, but, if memory serves, no differential equations or fancy stuff like Real Analysis (which doesn't actually involve numbers, in my experience). But it does help if you can see why you might want to integrate the distribution function (often a Gaussian) from here to infinity, and can go look up the resulting error function when you need it. And if you have the notion of related rates under your belt, so that propagation of errors makes sense (as opposed to being magic),

"And this ... is a piece of Synergy." --Kellnerin

[ Parent ]
going further off of this comment. by gzt (2.00 / 0) #4 Thu Jul 15, 2010 at 12:19:08 PM EST
It depends a lot on what sort of stats problems you're doing and what sort of data sets you're working with. Generally, and you might not like this, you'd get a practical exposure to "working with real data" in whatever field you're working at in upper-level courses in your field (depending on your field, that will either mean upper-level undergraduate or beginning post-graduate). If you can't actually do it in the field like that, what might be your best bet is to read papers (ugh) where you can get the raw data and run your own analysis of it, trying to replicate what they did (ugh).

Otherwise, there's a terribly practical guide to doing this stuff with social sciences data called The Logic of Causal Order by James Davis. He runs NORC and I had a class with him. He wears bow ties and suspenders. The monograph really helps you think about how you must run your regression analysis, and the thinking part is the most important of them all.

I have this same problem, in that I have a minimal statistical education and a very strong maths education, and people sometimes want me to do statistical analysis of crap here because I'm the only person resembling a statistician for miles. I mostly figure it out.

[ Parent ]
note: monograph does not have examples, really by gzt (2.00 / 0) #5 Thu Jul 15, 2010 at 12:21:36 PM EST
just a thoroughly helpful method for thinking about how to go about the whole thing of statistical analysis before you even start plugging numbers in. Once that's done, it becomes a lot simpler to do the analysis, though there's still some "technique" to it.

[ Parent ]
if you want to look at by sasquatchan (2.00 / 0) #3 Thu Jul 15, 2010 at 12:17:08 PM EST
random number generation, monte carlo simulations and what not, I suggest discrete event simulation. (I'm in the ack's for the book ;)

[+] [P] Books that give hands-on examples of how to use statistics properly | 6 comments (6 topical, 0 hidden)