Statistics Workbook

June 25, 2021 12:12 AM

I'm working in the life sciences and infrequently have to engage with statistics, like ANOVA, ttests and general data analysis of High Throughput data. But due to the infrequency I always forget a lot of this stuff and have to relearn it. I once had a really great statistics workbook, it was among some 20 books I just grabbed from the statistics shelf from my library to see what works best for me. It was quite short on the explanatory text, but had some outstanding problems to work on, the kind that makes you really think about the stuff you just learned, and not just mechanically apply the same formula with different numbers. These questions were very well designed and the answers to them were lengthy, so that you really understood the solution. For me this is the best method for truly understanding something. I want to find such a book again, can you help me out?
posted by SweetLiesOfBokonon to Science & Nature (7 answers total)

wow, I kind of want this too. Did it stop at ANOVA/t-test stuff, or do you recall other methods in particular? The course I had out of Rawlings, Pantula, and Dickey's "Applied Regression Analysis" cemented some key concepts in all those least-squares kinds of analysis, but it's not short and I don't recall there being any high-throughput data in there.

I'd be willing to bet that there's a bunch of options in the course-notes-for-undergrad-applied-stat-classes out there on the web, too. Oftentimes the best applied stuff is "Statistics for ______"; fill in the blank with whatever practice you like: engineers, dummies, chemists, scientists...everybody!

Depending on what software you have, the documentation might have a nice mix of detail and application. Example from SAS documentation on Logistic regression "details" and "examples."
posted by adekllny at 6:08 AM on June 25, 2021

Best answer: I haven't worked my way through this one yet so I can't speak to the quality/nature of the problems, but The Humongous Book of Statistics Problems is a commonly recommended one.
posted by thebots at 9:29 AM on June 25, 2021 [1 favorite]

Response by poster: I work in R and the book was strictly concerned with classical statistics like distributions, probability and the various test statistics.
posted by SweetLiesOfBokonon at 12:54 PM on June 25, 2021

On the eee the old ones are the best ones front can I put in a plug for MJ Moroney's Facts from Figures? First published before (even) I was born, there are numerous editions of (blue) Pelican paperbacks out there. I pick them up at yardsales and give them to any student with mathy tendencies. Written before R, SPSS or Excel, indeed before calculators, FfF has to explain the principles because the calcs are all done with pencil and paper. Moroney obit.
posted by BobTheScientist at 1:04 PM on June 25, 2021 [3 favorites]

All of statistics is another book that might fit. I commonly see it recommended, but I'm not sure about whether it has good problems.
posted by Maxwell_Smart at 3:25 PM on June 25, 2021

Best answer: All of Statistics is great and has a lot of practice problems, but is relatively terse and fast-paced so you may have to supplement with something more chatty.

I would actually recommend Holmes and Huber, which you can even access online for free. It’s a little unorthodox in how it orders topics (for example, it starts with count based data and the Poisson distribution) but it’s for good reasons, and the text is explicitly geared towards biologists using modern tools as the name implies. I like the explanations a lot and it has R code and examples.
posted by en forme de poire at 8:56 AM on June 26, 2021

I have a copy of Introduction to Biostatistics by Sokal and Rohlf. Topics are explained adequately without overwhelming my non-numeric brain.
posted by unearthed at 4:50 PM on June 26, 2021