Summer semester 2023-2024, Thursdays, 17:40-19:10 (Room 237)
Rectorate Building, Wiejska 45-A, Bialystok
We will be meeting every week during the first half of the semester
Teacher: Marek J. Druzdzel
"Science is not science fiction. It accepts the tests of observation and experiment, acknowledges the supremacy of fact over wish or hope. The smallest experiment can crash to earth the most attractive theory." --- Herbert A. Simon
The course "Research Design" introduces doctoral students to a rich field of empirical (i.e. experimental) research and guides them through the meanders of problems related to learning the world.
The course is neither intended to teach you formal details of statistical procedures nor to make students experienced practitioners in specific areas design tools. This can and should be acquired in other specialized courses. The aim of this course is to help the participants in developing broad critical skills and the emphasis will be on the basic process of scientific inquiry. The lecture is designed to improve your ability to think about research questions, formally formulating them, and choosing ways of solving them through quantitative empirical research.
The approach taken in this course is somewhat unorthodox in comparison with what can be found in existing textbooks and courses on the topic of research design. We will start with the concepts of causality and causal graphs and how how they represent statistical independence. Causal graphs are close to directed probabilistic models, such as Bayesian networks, increasingly used in decision support systems. Understanding causal graphs will help you to gain insight into the structure of scientific experiments and understand what experiments are. Like any lecture on design of experiments, this one will cover basics of experimental design and topics that are directly related to it, such as identifying and articulating research problems, formulating testable hypotheses, measurements and data collection, artifacts occurring between the subject and the experimenter and their control, describing and displaying data, interpreting and drawing conclusions from data analysis, and reporting research results and their implications. The course will also cover less orthodox topics, namely discovery of causal structures from data and computer simulation.
Those of you who would like to play with the GeNIe, the program that I will use in class to explain causal graphs and to demonstrate causal discovery from data, can download it at the following address: https://www.bayesfusion.com/. Academic version of GeNIe is free for teaching and research purposes.
Grading will be based on presence and also a simple project consisting of a description of a scientific problem and a proposal for an experimental solution to it. Ideally, the selected research problem should be part of your dissertation. A project report of a maximum length of five pages must be submitted within a week of the end of the course.
Outline of the lecture content and meeting dates:
Class materials
Marek Druzdzel's teaching page
Marek Druzdzel's home page
marek@sis.pitt.edu /
Last update: 7 March 2024