HN Explorer

Ask HN: Any books that focus on understanding hypothesis testing?

2 points by BOOSTERHIDROGEN ↗ HN
including probability distributions, confidence intervals, p-values(hacking (https://xkcd.com/882/) ), sampling and methods of conducting experiments? would like a book that explains intuitively and directly instead of having me extract them from examples.

2 comments

[ 3.0 ms ] story [ 16.6 ms ] thread
mindcrime
Sounds like you want a book that would typically be titled "Design of Experiments" or "Experiment Design" or something similar. There are tons of them on the market[1][2][3][4], probably any mainstream one used for a university course on the subject would be adequate.

Here's a freely available book / lecture notes doc that covers some of the ground you're interested in.

http://www.stat.rice.edu/~dobelman/courses/texts/Notes.411_6...

[1]: https://www.amazon.com/Design-Analysis-Experiments-Springer-...

[2]: https://www.amazon.com/Design-Experiments-Mike-Peralta/dp/14...

[3]: https://www.amazon.com/Design-Experiments-Approach-Bradley-J...

[4]: https://www.amazon.com/Design-Analysis-Experiments-Douglas-M...

BOOSTERHIDROGEN
Let say I'm using statistics program like SPSS. I have 20 different datasets that explain 20 different variables.

The Scientific Method would be to propose a hypothesis, determine that the hypothesis would be supported is A was related to G, and then tell SPSS to run a crosstabulation of A and G.

Alas, the software can just as easily run a cross tabulation of all 380 possible combinations of variables. There is even a special command to do this. With p-Hacking I look through these 380 tables to find the one with the best correlation (let's say that is C and R). Then I try and guess up a theory that C being correlated to R supports.

I can't logically validate a hypothesis but only reject the opposite hypothesis.