Week 02 — [Philosophical concepts in data analysis]
Week: 2
UPDATE: Due to inclement weather, there will only be one session this week (Thursday).
Topic: Foundational philosophical questions
1 Background
The group that read Tredennick et al. (2021) will be giving their lecture.
Review the following topics if you don’t fully understand them:
- Inference
- Probability
- Statistical models
- Samples and Populations
- Bias and Variance
- Frequentist and Bayesian statistics
- Uncertainty in data analysis
You can use AI or a quick search to help you understand these concepts, but make sure you understand them somewhat
2 Learning Objectives
Articulate what statistical models are for and how they function as scientific tools
Explain how uncertainty enters data analysis and how it is handled in different inferential frameworks
Compare frequentist and Bayesian perspectives on statistical inference
Critically evaluate the assumptions underlying quantitative conclusions in published research
3 Readings
Please read this paper:
Ellison (2004) by Tuesday.
Again, a pretty complex topic. If you are not familiar with Bayesian statistics, focus on the philosophical side of things rather than the technical aspects.
And explore this paper by Thursday: Box (1976) (available at: https://www.jstor.org/stable/2286841?seq=1)
Box (1976) ‘Science and Statistics’ is VERY dense and had to read, so do not worry about every detail, and you can skim through it. For this class, your goal is to understand three ideas: (1) science as an iteration between theory and data, (2) models as deliberately false but useful, and (3) the need to check and revise models when they clash with data.
3.1 Session A:
3.1.1 Part 1: Lecture by Team B (~ 25 minutes of the class)
The team that did not give their lecture today, will be giving a lecture on choosing models.
3.1.2 Part 2: Paper discussion (~25 minutes of the class)
I will open with a short lecture on Ellison (2004) (15 minutes), then we will have a discussion (15 minutes).
The discussion will be guided by some questions I will write.
What does it mean for a model to be “useful” rather than “true”?
What kinds of uncertainty are explicitly modeled, and which are often ignored?
How do Bayesian and frequentist approaches differ in how they treat uncertainty?
How do these philosophical perspectives influence how we interpret results in applied research?
3.1.3 Part 2: Dataset/coding
We will start coding here. The objective will be to run a Bayesian example on Thursday
3.1.4 Exit ticket
What was the most challenging thing about today?
3.2 Session B
Methods workshop — hands-on coding Session B has been cancelled