Course 3
COURSE 3 · MATHEMATICAL & STATISTICAL METHODS
Contents
Every tutorial in the course, topic by topic. Within a topic the tutorials are listed alphabetically.
← Back to Course 3 — Mathematical & Statistical Methods
Topic 1 · Linear Algebra & Matrices
Eigenvalues, SVD, positive definite matrices, Perron-Frobenius, graph Laplacians. Topic page →
| Title | Description |
|---|---|
| Eigenvalue methods for repeated game analysis | Use eigenvalue decomposition to compute stationary distributions, convergence rates, and long-run cooperation levels for Markov strategies in the repeated Prisoner’s Dilemma, including Tit-for-Tat, Grim Trigger, and stochastic strategies. |
| LP duality and zero-sum game solutions | Explore the formal connection between linear programming duality and the minimax theorem for zero-sum games, implementing both primal and dual formulations to solve a 3x3 game. |
| Markov Chains for Strategy Dynamics in Game Theory | Model strategy updating as a Markov chain in populations playing 2x2 games, computing transition matrices, stationary distributions, and mixing times for imitation dynamics and mutation-selection processes. |
| Matrix games — linear algebra foundations for game theory | Express strategic games as matrices, compute mixed-strategy Nash equilibria using linear algebra operations including eigenvalues, null spaces, and linear programming, and build a foundation for numerical game theory in R. |
| Singular value decomposition for game analysis | Apply SVD to analyze payoff matrices, identify dominant strategic dimensions, construct low-rank game approximations, and connect matrix rank to mixed strategy structure. |
Topic 2 · Optimization & Numerical Methods
Convex optimisation, LP, dynamic programming, gradient methods, fixed-point. Topic page →
| Title | Description |
|---|---|
| Genetic algorithms for equilibrium computation | Apply genetic algorithms to find Nash equilibria in bimatrix games by encoding mixed strategies as chromosomes, using negative regret as fitness, and designing crossover and mutation operators on probability simplices. |
| Gradient Descent Methods for Finding Nash Equilibria | Implement gradient-based optimisation algorithms for finding Nash equilibria in continuous games, including best-response dynamics, the Nikaido-Isoda function approach, and simultaneous gradient descent with convergence analysis in R. |
| Interior Point Methods for Solving Game-Theoretic Optimization Problems | Implement barrier methods for the linear programming formulation of zero-sum games, visualise the central path and barrier parameter convergence, and apply log-barrier techniques to compute correlated equilibria. |
| Simulated annealing for equilibrium computation | Apply simulated annealing to find Nash equilibria in complex games, comparing temperature schedules and neighborhood structures with gradient-based methods. |
| Support enumeration algorithm for Nash equilibria | Implement the support enumeration method to find isolated equilibria from nonsingular equal-size supports in bimatrix games by solving indifference conditions for each support pair, with applications to games with multiple equilibria. |
| The Lemke-Howson algorithm for computing Nash equilibria | Implement the Lemke-Howson pivoting algorithm for bimatrix games in R, trace the complementary path from an artificial equilibrium to a Nash equilibrium, and visualise the pivot steps on the best-response polytope. |
Topic 3 · Information Theory
Entropy, mutual information, KL divergence, channel capacity, rate-distortion. Topic page →
| Title | Description |
|---|---|
| Cheap talk and strategic communication | The Crawford-Sobel (1982) cheap talk model: how costless, non-verifiable messages can transmit information when preferences are misaligned, with partition equilibria, information loss as a function of bias, and connections to costly signalling games. |
| Entropy and the value of information in strategic games | Apply Shannon entropy to game-theoretic information sets, quantify uncertainty reduction, and compare the value of perfect versus imperfect information using R implementations. |
| Maximum Entropy Correlated Equilibria | Compute the maximum entropy correlated equilibrium of a game as a constrained optimisation problem in R, comparing it with welfare-maximising CE and analysing mutual information between signals and actions. |
| Mutual information in signaling games | Quantify information transmission in signaling games using Shannon mutual information. Compare pooling versus separating equilibria by their channel capacity and information content. |
| Rational inattention: information costs and strategic behaviour | Model agents who face Shannon mutual information costs for acquiring information about the state, showing how limited attention leads to sticky behaviour, stochastic choice, and deviations from full-information equilibrium. |
| Value of information in strategic games | Compute the value of perfect and imperfect information in decision problems and games, showing that more information always helps a single agent but can hurt in strategic interaction. |
Topic 4 · Statistical Foundations
Probability, inference, hypothesis testing, regression, nonparametrics. Topic page →
| Title | Description |
|---|---|
| Bayesian inference for game-theoretic parameters | Estimate game parameters including payoffs, type distributions, and rationality levels from observed play using Bayesian methods with quantal response equilibrium likelihoods and grid-based posterior computation. |
| Bootstrap inference for game-theoretic quantities | Apply bootstrap resampling to construct confidence intervals for mixed-strategy frequency estimates and Shapley values, with coverage simulations comparing percentile and BCa methods. |
| Bootstrap methods for game-theoretic inference | Apply bootstrap resampling to construct confidence intervals for equilibrium predictions, including nonparametric bootstrap for payoff estimates, parametric bootstrap for equilibrium strategies, and bootstrap hypothesis tests for equilibrium play. |
| Hypothesis Testing in Strategic Environments | Frame statistical hypothesis testing as a game between researcher and Nature, implement permutation tests for Nash equilibrium deviations, and demonstrate multiple testing corrections across games. |
| Hypothesis testing as a game between nature and the statistician | Reframe classical hypothesis testing as a two-player minimax game in which nature chooses the true state and the statistician selects a decision rule. Derive the Neyman-Pearson lemma as the minimax optimal strategy and explore power analysis as strategic optimisation. |
| Maximum Likelihood Estimation of Game-Theoretic Models | Estimate the quantal response equilibrium (QRE) rationality parameter from experimental choice data using maximum likelihood, visualize the log-likelihood surface, and compare QRE with Nash equilibrium predictions. |
Topic 5 · Bayesian Methods
Conjugate priors, MCMC, Stan, hierarchical models, Bayesian computation. Topic page →
| Title | Description |
|---|---|
| Bayesian Estimation in Common-Value Auctions and the Winner’s Curse | Model common-value auctions with noisy signals, derive optimal bids accounting for the winner’s curse, and simulate the devastating effect of naive bidding on expected profits. |
| Bayesian games — strategic interaction under incomplete information | Formalise games of incomplete information in R, compute Bayes-Nash equilibria for type-dependent strategies, and visualize how private information shapes equilibrium behaviour. |
| Bayesian persuasion and information design | Implement the Kamenica-Gentzkow (2011) Bayesian persuasion model in R, computing optimal signal structures via concavification for the binary-state prosecutor-judge example. |
| Global games and coordination under private information | Implement the Carlsson-van Damme and Morris-Shin global games framework in R, showing how private noisy signals about fundamentals resolve equilibrium multiplicity in coordination games through threshold strategies. |
| Optimal reserve prices in auctions — Myerson’s revenue-maximising design | Derive Myerson’s optimal reserve price for sealed-bid auctions, implement revenue comparisons across uniform and non-uniform value distributions, and visualize the revenue gains from optimal auction design in R. |
Topic 6 · Causal Inference
DAGs, potential outcomes, DiD, IV, RDD, synthetic control. Topic page →
| Title | Description |
|---|---|
| Difference-in-differences with strategic agents | Applying the difference-in-differences estimator to settings where treatment alters a strategic equilibrium, demonstrating how agents’ endogenous responses can violate the parallel trends assumption and bias causal estimates. |
| Instrumental variables for causal effects in strategic settings | Use instrumental variable methods to estimate causal effects when strategic interaction creates endogeneity, implement two-stage least squares (2SLS) in R, and apply the approach to a game-theoretic model of market entry. |
| Instrumental variables in strategic settings | An application of instrumental variables methods to estimate causal effects in games where strategic actions are endogenous, covering simultaneity bias, instrument validity, two-stage least squares with simulated game data, and weak instruments diagnostics. |
| Regression discontinuity in strategic settings | Regression discontinuity design applied to games: how a regulatory threshold creates a discontinuity in strategic incentives, with local polynomial estimation of the treatment effect and a McCrary density test to detect strategic manipulation of the running variable. |
| Synthetic Control Methods for Strategic Policy Interventions | Apply the synthetic control method to estimate the causal effect of a competition policy change in a strategic environment, using constrained optimisation to construct counterfactual outcomes from untreated regions. |
Topic 7 · Time Series & Econometrics
ARIMA, VAR, cointegration, GARCH, structural breaks, panel data. Topic page →
| Title | Description |
|---|---|
| Cointegration analysis of strategic long-run relationships | Test for cointegration between strategic variables such as rival firm prices and competing military budgets using Engle-Granger and Johansen methods, and estimate vector error correction models for strategic adjustment dynamics. |
| Granger causality in strategic interaction data | Apply Granger causality tests to detect strategic interdependence in time-series data, distinguishing simultaneous Cournot play from sequential Stackelberg leadership in simulated oligopoly settings. |
| Modelling strategic interaction with VAR models | Use vector autoregression to detect and model strategic interaction in time series data, with arms race dynamics as the motivating example. Granger causality tests reveal whether one actor’s past behaviour predicts another’s future choices, formalising the notion of strategic influence. |
| Structural Estimation of Oligopoly Conduct Parameters | Estimate the Bresnahan-Lau conjectural variations parameter from simulated Cournot oligopoly data using 2SLS/IV estimation, and compare conduct regimes from perfect competition to full collusion. |
| VAR Models for Strategic Interaction Dynamics | Vector autoregression models capture dynamic strategic interdependence in oligopoly markets, revealing how firms’ quantities today depend on rivals’ past decisions through impulse response functions and Granger causality tests. |
Topic 8 · Network Science
Random graphs, centrality, community detection, epidemics, igraph/tidygraph. Topic page →
| Title | Description |
|---|---|
| Centrality measures for game-theoretic network analysis | Compute degree, betweenness, and eigenvector centrality for strategic networks using igraph in R, and visualise how different centrality measures reveal distinct power structures in game-theoretic settings. |
| Community Detection as Coalition Formation | Discovering network communities through the lens of cooperative game theory — modelling dense subgroups as coalitions, implementing modularity-based detection, and analysing Nash stability of the resulting partitions. |
| Diffusion and cascades on networks | Simulate information cascades using the linear threshold and independent cascade models on random networks, demonstrating tipping points and the effect of network structure on cascade size. |
| Power-law networks and strategic behaviour | Analyse how scale-free degree distributions in Barabasi-Albert networks affect coordination game outcomes, hub influence on equilibrium selection, and targeted vs random intervention strategies using R simulations. |
| Strategic network interdiction: max-flow, min-cut, and Stackelberg games | Model network interdiction as a two-player zero-sum game where an attacker removes edges to minimise maximum flow while a defender protects critical links. Explore the resilience-efficiency trade-off. |