Lecture Notes for Advanced Quantitative Political Methodology

These are materials from class, previous years, and section (all with videos, slides, R code, and other materials).   When I lecture, I go through as much material as possible at each lecture, subject to the constraint that everyone follows what I'm doing. The speed at which I go is therefore dependent on the composition of each year's class and the questions that arise. As such, the slides below are not broken up into distinct weeks (I seem to go through roughly 15-20 pages in a weekly session lasting almost 2 hours, but issues and topics not represented here are covered most weeks).  

In class, in addition to lecturing, we will use LearningCatalytics (which I helped develop), along with some other demos and collaborative learning tools.

Separate PDF versions of the slides appear for teaching (i.e,. with pauses, etc.) and for (color) printing. R code is provided for each section. I update this material almost continuously while I teach. I'd very much appreciate if you would contact me with any comments, corrections, or suggestions.

Videos for 2014 lecture and section will be available in the sections at the right, and also here.  

Videos from previous years are available here:20132012. Access may require being logged in via your Harvard ID.

The Basics

Theories of Inference

Alternative theories of inference (Bayes, likelihood, Neyman-Pearson hypothesis testing, etc.)

2014 Lecture

  Lecture 4 Video (2 hrs)
  Lecture 3 Video (2 hrs)

  
Slides used in class
  Slides for printing

2014 Section

  Section 3 Video (1 hr)
  Section 3 Slides
  Section 3 Code
  Section 4 Video (1 hr)
  Section 4 Slides
  Section 4 Code
  Section 5 Video (1 hr)
  Section 5 Slides
  Section 5 Code

2013 Lecture

 Lecture 3 Video (2 hrs)
 Lecture 4 Video (2 hrs)
 Lecture 5 Video (2 hrs)
 Slides used in class
 Slides for printing

2013 Section

 Section 3 Video (1 hr)
 Section 3 Slides
 Section 3 Slides (printable)
 Section 4 Slides
 Section 4 Video (1 hr)
 Computation Section Code
 Section 5 Video (1 hr)
 Section 5 Slides
 Section 5 Slides (Printable)

2012 Lecture

 Lecture 3 Video (2 hrs)
 Lecture 4 Video (2 hrs)
 Slides used in class
 Printable version of slides

2012 Section

 Section 2: Introduction to Likelihood
 Section 2: Slides
 Section 3: Optimization
 Section 3: Slides
 Section 4: Likelihood (cont.)
 Section 4: Slides

2010 Lecture (selections):

 Slides used in these videos
 Problem of Inference and Likelihood Theory of Inference (36:34)
 Bayesian Theory of Inference (22:14)
 Neyman-Pearson Hypothesis Testing (13:03)
 What is the Best Theory of Inference? (09:29)
 Likelihood of Stylized Normal (28:44)
 Hyperdimensional Space (11:04)
 Finding Maximum of Likelihood (05:15)
 Properties of Likelihood Maximum + Uncertainty (23:02)
 Standard Errors of Maximum Likelihood Estimates (29:20)
 Simulation for any Maximum Likelihood Model (08:37)
 Forecasting Presidential Elections - Part 1 (18:42)
 Forecasting Presidential Elections - Part 2 (21:01)
 Variance Function Models (08:05)

Single Equation Models

Binary variable models, interpretation and presentation via simulation, ordinal dependent variables, how do you know which model is better?, grouped binary variables, event counts, simple duration models and censoring. 

2014 Lecture

  Lecture 5 Video (2 hrs)
  Lecture 6 Video (2 hrs)
  Lecture 7 Video (2 hrs) 
  Lecture 8 Video (2 hrs) 
  Slides used in class
  Slides for printing
  Slides about robust standard errors

2014 Section

  Section 6 Video (1 hour)
  Section 6 Slides 
  Section 6 code and data
  Section 7 Video (1 hour)
  Section 7 Slides
   Section 7 code and data
  Section 8 Video (1 hour)
  Section 8 Slides 
  Section 8 code
  Section 9 Video (1 hour)

  Section 9 Slides 

  Section 9 code

2013 Lecture

 Lecture 5 Video (2 hrs)
 Lecture 6 Video (2 hrs)
 Lecture 7 Video (2 hrs)
  
Lecture 8 Video (2 hrs)
 Slides used in class
 Slides for printing

2013 Section

 Section 6 Video (1 hour) (Part 1 is Dichotomous Outcomes; Part 2 is Robust Standard Errors)
 Section 7 Video (1 hour)
 Section 8 Video (1 hour)
  
Section 9 Video (1 hour)
 Section 5 Slides
 Section 5 Slides (printable)
  
Section 6 Slides
  
Section 6 Slides (printable)
 Section 7 Slides
 
Section 8 Slides (printable)
 Section 9 Slides (printable)

2012 Lecture

 Lecture 5 Video (2 hrs)
 Lecture 6 Video (2 hrs)
 Slides used in class
 Printable version of slides
 Single Equation Models

2012 Section

 Section 6: Dichotomous Outcomes and Presenting Quantities of Interest
 Section 6: Slides
 Section 7
 Section 7: Slides
 Section 8: Count Models, Counterfactuals, Duration Models
  Section 8: Slides

2010 Lecture, selections assigned this year:

 smodelsp2010.pdf
 Logit Model - Part 1 (10:49)
 Logit Model - Part 2 (17:42)
 Alternative Interpretation of Logit Model (25:13)
 Presenting Statistical Results - Part 1 (11:19)
 Presenting Statistical Results - Part 2 (09:11)
 Simulating Quantities of Interest and Associated Uncertainty (38:44)
 Application of Clarify Package (06:18)
 Ordinal Dependent Variable Models (26:01)
 Model Validation - Part 1 (13:21)
 Model Validation - Part 2 (25:51)
 Grouped Uncorrelated Random Variables (20:20)
 Event Count Models (27:37)
 Duration Models: Exponential (17:38)

Causal Inference

Detecting Model Dependence

Sensitivity to parametric assumptions, revealing inferences too far from the data to have empirical answers, the curse of dimensionality, extrapolation, measures of distance from the data.

2014 Lecture

  Lecture 9 Video (2 hrs)
  
Slides used in class
  Printable version of slides 
  5modeldependencematching.r

2013 Lecture

  Lecture 9 Video (2 hrs)
  
Lecture 8 Video (2 hrs)
 Slides used in class
 Printable version of slides 
 5modeldependencematching.r

2012 Lecture

 Lecture 7 Video (2 hrs)
  Slides used in class
  Printable version of slides
 5modeldependencematching.r

2012 Section

Section
Section: Slides

2010 Lecture, selections assigned this year:

 Slides used in these videos
 Model Dependence (27:42)
 Example of Model Dependence (Doyle and Sambanis, APSR 2000) (13:13)

Research Designs

An overview of how key features of various observational and experimental research designs, and the designs themselves, reduce components of error in estimating causal effects.

2014 Lecture

  Lecture 10 video (2 hrs)
  Slides used in class

  Printable version of slides
  R Code

2013 Lecture

 Lecture 10 Video (2 hrs)
 Slides used in class
 Printable version of slides
 R Code

2013 Section

  Section11
  Section 11 Slides

2012 Section

 Section10
 Section 10 Slides

2010 Lecture, selections assigned this year:

 matchsetlkp2010.pdf
 Causal Quantities of Interest and Error Decomposition (44:35)
 Benefits of Major Research Designs (38:28)

Multiple Equation Models

Analysis Models

Identification, how multiple and single equation models differ, seemingly unrelated regression models, reciprocal causation, multiple equation reparametrization, multinomial choice models (multinomial probit, multinomial logit), independence of irrelevant alternatives, conditional logit.

2014 Lecture

  Lecture 11 Video (2 hrs)
  Slides used in class

  Printable version of slides

2014 Section

  Section 12 Video (1 hr)
  Section 12 Slides
  Section 12 Code

2013 Lecture

 Lecture 11 Video (2 hrs)
  Slides used in class
  Printable version of slides

2013 Section

 Section12
 Section 12 Slides 

2012 Lecture

 Lecture 10 Video (2 hrs)
 Slides used in class
 Printable version of slides

2012 Section

 Section11
 Section 11 Slides
 Section 11 R Code

2010 Lecture, selections assigned this year:

 mmodelsp2010.pdf
 Multiple Equation Models and SURM (27:17)
 Reciprocal Causation (27:15)

Text Analysis

Text Analysis: Statistical analysis when the observation is a text document of some kind.

Supervised Learning

Supervised Learning: Estimating the average in a category vs. individual classification, dealing with measurement error, extensions to an apparently unrelated application in epidemiology and public health.

2012 Lecture Notes: Slides [to print] 

Videos ( Printable slides to accompany videos )

 Quantities of Interest in Text Analysis and Representing Text as Numbers (22:27)
  Existing Statistical Approaches to Estimating Proportions in Categories (10:22)
 Using Misclassification Rates to Correct Proportions (12:55)
  Verbal Autopsy Methods (07:18)

Unsupervised Learning

Unsupervised Learning: fully automated vs computer assisted cluster analysis; why humans are incapable of learning from text as well as computers & why computers can't do it without humans.

2012 Lecture Notes: Slides (no 'to print' version)

Videos ( Printable slides to accompany videos )

 New Strategy for Cluster Analysis (28:22)
  Applications of Cluster Analysis (19:33)

Time Series Fundamentals

The basics of time series models.

Lecture Notes: Slides [to print


Rare Events

Time Invariant Models

Rare events, relationship to logistic regression, classic case-control research designs (how and why to select on the dependent variable), robust bayesian analysis, reporting standards in case-control studies, ReLogit software.

Lecture Notes: Slides [to print]

Time Varying Models

Time varying quantities of interest; understanding hazard rates; exponential, Weibull, and Cox Proportional hazard duration models, density case control models.

Lecture Notes: Slides [to print]

Ecological Inference

Making inferences about individual behavior from aggregate, group-level data. History of the ecological inference problem, Goodman's model, the Davis-Duncan bounds, King's EI model, what can go wrong and what to do about it, and model extensions.

Lecture Notes: Slides [to print]