(2) This is an introduction to statistics for beginning grads or adv. undergrads. I first had the idea to write it when I was a graduate student in psychology at Berkeley, and I actually wrote the bulk of it after changing fields, while I was doing a PhD in population genetics.
-
-
Prikaži ovu nit
-
(3) At Berkeley I already knew I liked statistics and had a little math background, so I took courses from the stat dept (eventually doing an MA). My dear cohort-mates took statistics courses in the psych dept. Our experiences were totally different.
Prikaži ovu nit -
(4) My friends learned about a lot of different methods in their courses. At the end of their first year, they could run a bunch of different kinds of regressions and ANOVAs and interpret the output. It was a quick way to become familiar with the methods they'd see most often.
Prikaži ovu nit -
(5) In contrast, I learned few actual statistical methods in my first year, but I was taught in a way that let me understand the methods I did learn in a fairly deep way---I could simulate and/or do a little math to understand their properties.
Prikaži ovu nit -
(6) Even though I'm no smarter than any of my former cohort mates, and they arguably had more relevant methods training than I did, I quickly became the go-to person for statistics questions, because the training I got was way more adaptable to new situations.
Prikaži ovu nit -
(7) I started to think (as many others have argued) that the teach-a-bunch-of-methods approach is actually a broken model. I can see that it might have worked well in the past, but these days, empirical researchers have to wrestle creatively with lots of different kinds of data.
Prikaži ovu nit -
(8) I thought of my cohort-mates---super smart, motivated people who just hadn't taken much math or done much programming---and wondered how I'd give them an experience like I got in my first year at Berkeley. This book is my attempt at an answer.
Prikaži ovu nit -
(9) The basics of my answer are: (a) focus on one statistical method (I use simple linear regression, (b) take the math seriously but still be gentle when possible, and (c) rely heavily on simulations.
Prikaži ovu nit -
(10) The book starts by introducing some motivating questions and orienting the reader to R. After that, we consider the least-squares line without any probabilistic framing. Then there are a couple of chapters on probability.
Prikaži ovu nit -
(11) The second half of the book has the statistics. It starts with a chapter on point estimation, and then a chapter on interval estimation and hypothesis testing.
Prikaži ovu nit -
(12) Then the part I like best: we do simple linear regression three times, adding assumptions as we go. The first time we assume little and proceed with method of moments, bootstrapping, permutation. Then we do the standard normal maximum-likelihood approach, and then Bayes.
Prikaži ovu nit -
(13) The final chapter considers the checking and relaxing of assumptions, using it as an opportunity to sketch some related approaches (multiple regression, generalized linear models, mixed models).
Prikaži ovu nit -
(14) There is a github repo with all the code and exercise solutions, too, as well as an R package that implements all the functions in the book:https://github.com/mdedge/stfs
Prikaži ovu nit -
(15) People who knew me well while I was doing the bulk of the writing know that this has been a labor of love over many years, and I'm really excited that it'll be out soon.
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.