Introduction to Bayesian Data Analysis

1. 00:00Hi, my name is Shravan Vasishth. I'm a professor of linguistics at the University of Potsdam in
2. 00:05Germany. And I'm going to give you a short introduction to this "Introduction to Bayesian
3. 00:11Data Analysis", which is an open access course that will be provided at the portal openHPI.de
4. 00:20of the Hasso Plattner Institute. This course is free, and I'm going to tell you a little bit
5. 00:25about what's going to be in this course and what you can expect to learn from it. So what
6. 00:32we will do in this course is that we will study the foundational ideas behind Bayesian
7. 00:36methodology in data analysis, and we will run for four weeks.
8. 00:41This course will run for four weeks. And there are short lectures followed by quizzes
9. 00:45and exercises that you can do. And if you want a final certificate of completion, then
10. 00:51you can do a final exam.
11. 00:54The details for that will of course, be on the openHPI website. This course is designed
12. 01:02for a specific audience. So this course is appropriate for people
13. 01:08who already know statistical data analysis a little bit, that means you should
14. 01:13have done a little bit of R programming. You should have done some statistical data
15. 01:18analysis, for example, T-Tests or linear models, maybe even linear mixed models. But
16. 01:22if you're a complete newcomer to data analysis and you don't know any R, then this
17. 01:28course is probably not appropriate for you. I'm also assuming in this course that you
18. 01:34remember basic arithmetic operations - addition, subtraction, multiplication,
19. 01:38division from school, some very basic set theory, really elementary stuff, and just the
20. 01:45basics of probability theory, for example, the sum and product rule. And maybe you've
21. 01:49seen conditional probability in the past, even though you may not have used it in
22. 01:54practical settings. What I'm going to do in this course is to teach you modern
23. 01:59computational tools for doing Bayesian analysis. So you should want to
24. 02:05learn about that. So you should be interested in wanting to know about these
25. 02:08latest probabilistic programming languages that are available nowadays. One very
26. 02:13important thing that I assume in this course is what I call "a can do mindset". And what I
27. 02:19mean by that is that you need to have the ability to research questions that you don't
28. 02:25have answers to, that you may not have understood in the course of the
29. 02:29lecture, which need a little bit more detail that is missing from the lecture. So you
30. 02:34should be able to look this up yourself and to work on this yourself. And I've
31. 02:38actually prepared a little video and also a short blog post that spells out what I mean
32. 02:44by this. This is a very important attitude that you need to bring to the study of
33. 02:49anything that is technical, not just statistics. If you're doing anything else,
34. 02:53it's the same issue. You have to have this attitude of being able to unpack problems,
35. 02:58simplify them and to deal with them in a way that gives you a better understanding of the
36. 03:04material. So that's what I mean by I can do attitude, can do mindset. So you
37. 03:09should look at this, this lecture, this little video recording and this blog post.
38. 03:16What are we going to do in this course? Well, we're going to look at foundational ideas
39. 03:21about probability distributions and random variables, really basic things. And
40. 03:26then we will look at Bayes' rule and we will use Bayes' rule in practical applications
41. 03:31using analytical examples. So just paper pencil examples, and then I will switch to
42. 03:36doing more computational Bayes' that means I'll be using software packages, in
43. 03:41particular, BRMS, which is a front end to the probabilistic programming language, Stan. I'll
44. 03:46be using this package to demonstrate how you can fit more complex models such as linear
45. 03:52models and the end hierarchical models as well. So although we won't actually
46. 04:00get our hands dirty with the Stan programming language, we will be using a
47. 04:04front end called brms that uses the Stan language. And this should prepare you for the
48. 04:10future when you want to write customized models in the Stan language. So this is kind
49. 04:14of an entry into the world of probabilistic programming using this very powerful new
50. 04:20language, relatively new language called Stan. I will focus in the
51. 04:25computational part of the course. I will focus on regression modeling, linear models,
52. 04:29and maybe a little bit of hierarchical modeling. Not a lot, but these are the
53. 04:34foundational ideas that you need to build more complex models. So that's why this is
54. 04:38such an important topic in my view. So as I mentioned, you can, after finishing this
55. 04:45course, look at more advanced topics in Bayesian modeling and you'll be able to do a
56. 04:49lot more complex things that require this foundation that I'm going to teach. I will be
57. 04:56using a textbook that we have written. This book is available online and it will remain
58. 05:00online.
59. 05:01It will be published very soon with CRC Press. But the online version of the book
60. 05:07will be accessible to you at all times. And in this course at the Hasso Plattner Institute
61. 05:12portal, I will provide a PDF of this textbook so that you can read it offline as
62. 05:18well. Ok, so my advice to you would be to use this book as you're watching the
63. 05:25lectures, after you watch one week's lectures, it's a good idea to actually read
64. 05:30the lecture notes because they solidify the ideas that I present in short lectures. They
65. 05:36expand on those ideas in the lecture notes, so you should look at those carefully. And
66. 05:42once you've finished this course, I would advise you to take a look at the rest of the
67. 05:46chapters in the book, because these are the chapters that give you access to the more
68. 05:51complex models that are available in Bayesian methodology using the Stan
69. 05:56probabilistic programming language. Now, in order to prepare for this course, what you
70. 06:01should do is you should install R and RStudio, if you haven't done that already, you
71. 06:05should also install all the libraries that are mentioned in the introduction in the
72. 06:09textbook. So I will provide the installation code and you should also
73. 06:15familiarize yourself with the R markdown framework in order to write code in a
74. 06:22reproducible way. This is a very important thing, a very important skill that you should
75. 06:26acquire as a data analyst to be able to produce reproducible code. But this is
76. 06:32optional. You of course, don't need R markdown for this course. It's just my
77. 06:37suggestion that you will do well if you actually learn how to use R markdown for
78. 06:42your own data analysis problems. In closing, I want to mention that this course is
79. 06:48partially funded by a collaborative research centre that runs at Potsdam in the
80. 06:53linguistics and psychology departments in the Cognitive Science Department. This is called
81. 06:58Sonderforschungsbereiche in German, and the title of this collaborative research centre is The
82. 07:03Limits of Variability in Language. So I look forward to seeing you in this course and I
83. 07:09hope that you have fun.

To enable the transcript, please select a language in the video player settings menu.