Machine Learning for Psychology and Social Sciences [ML for Psychology]

Dozent/in:

Alexander Pastukhov

Angaben:

Seminar, 2 SWS, ECTS: 3

Termine:

Do, 10:00 - 12:00, M3N/-1.19

Voraussetzungen / Organisatorisches:

Basic knowledge of Python or R would be helpful.

Inhalt:

Psychology and social sciences rely statistical analysis to infer and describe causal effects that guide our behavior and determine our actions and responses. Proliferation of computers and smartphones, as well as a widespread access to internet made collection large datasets possible. In addition, movement towards open data in science means that there are many data sets both in panel databanks such as NEPS, SOEP, or ZPID, and at online repositories such as OSF or GitHub that can be analyzed and used to guide research and experimental design. However, the sheer amount of data makes using classic statistical methods cumbersome. The aim of the seminar is show how modern machine learning methods can be used to prescreen and analyze such big data sets. It will cover both the basic theory of methodology (math will be used very sparingly) and practical use of methods in Python and R (both free and open source systems). The material will cover a broad variety of supervised and unsupervised machine learning methods, including linear and logistic regression (which should be familiar from statistics), support-vector machines, tree-based approaches, cluster analysis, modeless analysis via nearest neighbor, deep neural networks, etc. A particular focus will be on application of these methods to "typical" social sciences / psychology data.

Empfohlene Literatur:

Hands-On Machine Learning with Scikit-Learn and TensorFlow by Aurélien Géron ISBN: 9781491962299

Python for social and experimental psychology II [Python for Psychology]

Dozent/in:

Alexander Pastukhov

Angaben:

Seminar, 2 SWS, ECTS: 3

Termine:

Do, 8:00 - 10:00, M3N/-1.19

Voraussetzungen / Organisatorisches:

Python for social and experimental psychology course or solid background in Python basics: variables, control structures (for and while loops, if conditional statements), lists/dictionaries, PsychoPy basics.

Inhalt:

The second part of the "Python for social and experimental psychology" course that covers advanced topics such as object-oriented programming the pythonic way, working with exceptions, use of iterators/generators for concise code, coroutines, use of scientific libraries (numpy, pandas), online programming via OTree system, etc. We we still be writing games (because psychological experiments are merely boring games).

Statistical Rethinking II [Statistical Rethinking]

Dozent/in:

Alexander Pastukhov

Angaben:

Sonstige Lehrveranstaltung, 2 SWS, ECTS: 3

Termine:

Mi, 14:00 - 16:00, M3N/-1.19

Voraussetzungen / Organisatorisches:

Some Bachelor level knowledge of statistics and R is beneficial, but no prior knowledge beyond high school algebra is required! For Ba / Ma Psychology only!

Inhalt:

Do you find statistics confusing and complicated? Do you want to improve and better understand your analysis? Do you want to find out that you are already a Bayesian statistician? Then this seminar is for you.

Learning Goals: In this seminar, you will learn how to build a statistical model from the ground up with the goal of being able to build a customized model for any statistical problem and analysis. After this course you will understand that a linear regression, a T-test, an ANOVA, or an ANOCOVA all refer to the same simple linear model that you can build yourself. The aim is to make sure that you will know exactly what your analysis does and why you are doing it in this way.

Course Method: This seminar assumes no prior knowledge on your part. We will start with a basic concept of probability-as-counting and proceed to understanding what statistical models are and how to build them. Over the course of the seminar, we will gradually move forward to more advanced topics learning how to handle various types of data, identify spurious associations, infer causality, evaluate models, or perform power analysis. Forming a book club we will read Statistical Rethinking by Richard McElrath. It is an excellent introductory statistics book that explain even most intimidating topics very clearly, links all seemingly discrepant topics together, and has plenty of examples in R. We will read one chapter every week and discuss the topics and questions during the seminar.

Empfohlene Literatur:

"Statistical Rethinking: A Bayesian Course with Examples in R and Stan" by Richard McElreath https://www.oreilly.com/library/view/statistical-rethinking/9781482253481/

Schlagwörter:

statistics, bayesian statistics