Estimate Heterogeneous Effects in Factorial Experiments Using Grouping and Sparsity
Calculate marginal effects
Plot a FactorHet object
Create the (Sparse) Design Matrix for Analysis
Create penalty matrix (list of F)
Deprecated Functions
Difference between AMEs in each group
Control for FactorHet estimation
Arguments for initializing FactorHet
Control for model-based optimization
Refit model using estimated sparsity patterns
Generic methods for FactorHet models
Estimate heterogeneous effects in factorial and conjoint experiments
Estimate heterogeneous treatment effects by individual or moderator
Compute association between moderators and group membership
Visualize the posterior by observed moderators
Predict after using FactorHet
Prepare Data
Rank of Matrix
Estimates heterogeneous effects in factorial (and conjoint) models. The methodology employs a Bayesian finite mixture of regularized logistic regressions, where moderators can affect each observation's probability of group membership and a sparsity-inducing prior fuses together levels of each factor while respecting ANOVA-style sum-to-zero constraints. Goplerud, Imai, and Pashley (2024) <doi:10.48550/ARXIV.2201.01357> provide further details.