# Very specific kind of textbook needed

I will (if I can find a good text) be teaching very basic stats to prisoners at a local med/high security prison.

Because of the conditions imposed I will have only fifteen 1.5 hours classes and I want to be able to get them through the very basics of definitions, distributions, probability and the concepts of testing.

They don't have access to PCs but can use hand calculators. In a run up to this class, I've done some probability stuff and prepared lots of handouts but that is a Herculean task.

All of these guys have done high school math. They are generally not educated but very smart, very intuitive - and grateful as hell to be treated as human beings. The class ranges from guys with 3 or 4 years to guys with life sentence plus 30 years; everyone is trying to improve themselves and they are great students.

This year we have some $to buy texts (The class size is about 15; I have been buying older versions of books myself but that starts to get costly for me and the books I've gotten haven't functioned very well.) Suggestions will be welcome. Something a bit above 'Statistics for Dummies' and well below Freedman, Pisani, Purves would be great. (I used 'Statistics for Dummies' one cycle and I hated it because I try really hard to treat these people the way I would treat any student - with respect and concern - and I don't want them to get the impression that I think they're dummies.) • @Gus It's a good bet that any book with the phrase "random processes" in it anywhere would be above the level of Freedman, Pisani, and Purves! – whuber Jul 5, 2016 at 21:12 • How are the students going to get their homework done? They can't post their assignments to Cross Validated for board members to do as can regular college students. Jul 5, 2016 at 21:29 • One thing people often overlook is that there are statistical procedures designed to be done by hand. This is in fact where many of Tukey's displays (stem-and-leaf displays, boxplots etc) and even tests (such as the Tukey Quick test and the Olmstead-Tukey test) originate; or Tukey's 3 group line, for example. Many rank-based tests can fall into this category, especially since the statistic may in some cases be found readily-enough by hand in samples that are nevertheless large enough to use asymptotic approximations - consider the sign test or the Kendall correlation (/Mann-Kendall trend test) Jul 5, 2016 at 21:52 • In small samples one could add the Theil-Sen regression line (which itself is connected to Kendall correlation). Indeed for small-to-moderate samples, signed rank and rank-sum tests are quite doable, for example, and these can lead to discussion of permutation tests, which in small samples reduce to the problem of counting up to no more than$\alpha N$where$N$is the number of sample-permutations under consideration; when$N$is small, this can be easily done by hand. Jul 5, 2016 at 21:54 • Many of these things I mentioned are distribution-free* or at least robust, and so concerns about distribution shape can be much less of an issue. Some of the more usual intro-stats curriculum can be done by hand of course -- but tends to be considerably more arithmetic-heavy. Sorting and counting are easy by comparison.$\:\$ * for continuous distributions at least. Jul 5, 2016 at 22:02