Week 1 |
Intro
Probability review
Maximum likelihood, R code
Bayesian inference intro (to be continued), R code
|
Reading: Review conditional probability. Shalizi Ch. 5
Studies mentioned in class: sex ratio at birth. Cohen et al., Insult, Aggression, and the Southern Culture of Honor: An "Experimental Ethnography"
|
Week 2 |
Bayesian inference intro (cont'd), R code
Introduction to statistical inference R code
The Truth About Linear Regression
|
Reading: Shalizi Ch. 2 and Ch. 5
Reading: When to control for lurking variables? Harvard admissions, the gender wage gap
Video: Bayesian Inference about Unicorns
Reading: OpenIntro Statistics 4.3.4, 5.1-5.2.1
Reading: the American Statistical Association's statement on p-values. Unilever statement on q-tips: "People may use [Q-tips] for ear cleaning, but we instruct against it," said Stanton of Unilever. Andrew Gelman: "I've never in my professional life made a Type I error or a Type II error"
Just for fun: the Replication Crisis
Just for fun: the scandal around the Stanford Prison Experiment
Just for fun: Psychology journal bans P values (N.B., this did not catch on.)
Just for fun: the Princeton connection to Darwin's finches
|
Week 3 |
Hierarchical models
Hierarchical models case study: restaraunt chains
Causal inference
Code (radon): multi.Rmd (html). Data: srrs2.dat
Code (finches): finches.Rmd (html)
|
Reading: Shalizi, Chapters 1-3. Shalizi & Gelman, Philosophy and the Practice of Bayesian Inference. Gelman & Hill, Data Analysis Using Regression and Multilevel/Hierarchical Models, Ch. 12.
Shalizi, Chapters 21-24.
|
Week 4 |
Causal inference (cont'd).
Code: fake_shaq.R
Code: polls.R (polls.dta)
Code: fake_shaq.R
Linear classifiers. logreg2d.html (Rmd)
Precept: Precept 4 (Rmd)
|
Reading: Shalizi, Chapters 21-24.
Just for fun: the decision in SFFA vs. Harvard
Just for fun: Basketball skills and height, conditioned on being in the NBA
|
Week 5 |
Intro to k-Nearest Neighbors
Logistic Regression on high-dimensional datasets
Word embeddings
Precept: asbestos_causal.R, heart_causal_handout.R
|
Reading: Sen and Wasow, Race as a Bundle of Sticks:
Designs that Estimate Effects of Seemingly Immutable Characteristics
Reading: CIML Ch. 3, CIML Ch. 7
|
Week 6 |
Generative models
Handout: gaussian_cancer.R
Precept: Python and word embeddings
|
Reading: CIML Ch. 9
|
Week 7 |
Intro to neural networks
ImageNet demo
Demo: AlexNet
Gradient descent
Overfitting
Training neural networks
Intro to Convolutional Networks
Precept: nb.html
|
Reading: CIML, Ch. 10
Paper: Caliskan et al., Semantics derived automatically from language corpora contain human-like biases, Science 356, (2017).
Paper: Greenwald et al., Measuring Individual Differences in Implicit Cognition:
The Implicit Association Test, J. of Personality and Social Psychology Vol. 74, No. 6 (1998). Project implicit at Harvard. Can We Really Measure Implicit Bias? Maybe Not in the Chronicle of Higher Education.
Paper (seminar on Friday): Spirling and Rodriguez, Word Embeddings: What works, what doesn’t, and how to tell the difference for applied research
|
Week 8 |
Neural networks handout
Intro to Convolutional Networks (cont'd)
Precept: Intro to NumPy (prez.jpg),
Gradient Descent,
Intro to PyTorch (solution: linear regression with PyTorch)
|
Paper (advanced): Belkin, Reconciling modern machine-learning practice and the classical bias–variance trade-off, PNAS 116(32) (2019)
Paper (mentioned briefly): Dissecting racial bias in an algorithm used to manage the health of populations, Science Vol. 366, Issue 6464, pp. 447-453 (2019)
Reading: CIML, Ch. 10 (continue)
Reading (reference): SciPy Lecture Notes
Reading (reference): Deep Learning with PyTorch: A 60 Minute Blitz
|
Week 9 |
Recap: Training machine learning models with gradient descent
Overfitting (cont'd)
Maximum Likelihood with PyTorch
Classifying digits with PyTorch (mnist_all.mat)
For fun: The Bee Gees' How Deep is Your Love in PyTorch (source code)
Precept: Fitting neural networks in PyTorch (some solutions)
|
Reading: CIML, Ch. 10 (continue)
Reading: Deep Learning with PyTorch: A 60 Minute Blitz (continue)
|
Week 10 |
Transfer Learning and Unsupervsied Learning
Intro to RNNs — generating language
Machine Translation with RNN
RNN worksheet
Precept: Mini-Project 3
|
Reading: cs231n notes on transfer learning
Reading: Karpathy, The Unreasonable Effectiveness of Recurrent Neural Networks (2015)
Just for fun: Multilingual Neural Machine Translation and "Machine Interlingua"
|
Week 11 |
Fairness in Machine Learning
Fun with ConvNets + transfer learning
Presentations
|
Reading: The Measure and Mismeasure of Fairness: A Critical Review of Fair Machine Learning. NIPS 2017 tutorial on Fairness in Machine Learning (slides, video). The analysis by Corbett-Davies et al of the COMPAS dataset in a WaPo blog post.
Papers mentioned: Farid and Dressel, The accuracy, fairness, and limits of predicting recidivism. Mitchell et al, Model Cards for Model Reporting.
Paper on monkey brains and trasnfer learning: Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition
Papers: Neural Style Transfer from the Bethge lab at the University of Tübingen.
Just for fun: Deep Dream grocery trip
Just for fun: the Matthew Effect
|
Week 12 |
To what extent is published research reproducible?
Presentations
|
Papers on reproducible science and false discovery rates:
- John Ioannidis, Why Most Published Research Findings Are False (PLoS Medicine, 2005)
- Jager and Leek, An estimate of the science-wise false discovery rate and application to the top medical literature (Biostatistics, 2014). Also see the discussion by several authors
- Open Science Collaboration, Estimating the reproducibility of psychological science
(Science, 2015)
- Simmons et al, False-Positive Psychology: Undisclosed
Flexibility in Data Collection and Analysis Allows Presenting Anything as Significant (Psychological Science, 2011). (See also The Garden of Forking Paths: Why multiple comparisons can be a problem, even when there is no “fishing expedition” or “p-hacking” and the research hypothesis was posited ahead of time, which we discussed before)
- Brad DeLong and Kevin Lang, Are All Economic Hypotheses False? (Journal of Political Economy, 1992)
Just for fun: was the Stanford prison experiment a real experiment?
|