Please do not reproduce any part of these notes without the author’s permission. Copyright © 2019

This course trains students to use statistical models for forecasting societal, mainly political, outcomes. The students learn how to use Machine Learning and Data Mining algorithms to explore topics such as measuring the extent of partisan polarization, predicting electoral outcomes, predicting local violence, analyzing the trend of interstate war, and forecasting civil war. Subjects to be covered include understanding the differences and similarities between Correlation Analysis, Causal Inference, and Forecasting Principles; Naive Bayes; k-Nearest Neighbors (KNN); Regularized Linear Regression (Lasso, Ridge, eNet); forecasting using Maximum Likelihood Estimation (MLE); Trees methods; Clustering; and Dimension Reduction.

Correlation Analysis vs. Causal Inference vs. Prediction/Forecasting

• A few of R Markdown cheetsheets: here and here. There are small differences between GitHub Markdown and R Markdown. Here is a cheetsheet for GitHub Markdown.

• If you need to refrsh your mind on statistics and mathemetics for this course, you may go over the handout of Political Science PhD Math Camp I prepared and taught at Arizona State Univeristy.

• Slides

Discrete Probability Review and Naive Bayes

k-Narest Neighbor (K-NN)

Linear Models and Regularization

Maximum Likelihood Estimation

Count Models: Poisson, Negative Binomial, and Zero-Inflation Estimation

Classification Metrics

Trees and Random Forest

• You might need to use Parallel computation in this course. Here is a Brief Introduction to Parallel Computation in R by Parallel.