These articles, undertaken with various collaborators, are in various stages of progress. The links below are not to the papers themselves but rather point to related ideas, just to give some sense of the background behind some of these projects.

• An agent-based model of coalitions and political instability (see here)

• Inference for a ratio when the denominator could be positive or negative (see here)

• Better than difference in differences (see here)

• Problems with formulations of risk aversion (see here and here)

• Red state blue state now (see here and here)

• Putting race, ethnicity, nationality, religion, and language identities on a common scale

• Poisson or negative binomial regression (see here)

• Interrogating the metaphor of cargo cult science

• Regression disconinuity analysis: What went wrong and how to do better (see here)

• No evidence for a role of maternal adiposity in sex ratio: An example of a hopelessly-noisy statistical analysis in epidemiology

• Model-based winsorizing

• Graphing uncertainty in election forecasting (see here)

• Understanding state-level polling errors

• Effective sample size of the prior distribution (see here)

• Static sensitivity analysis (see here)

• Why it doesn’t make sense in general to form confidence intervals by inverting hypothesis tests (see here)

• When MRP goes bad: Diagnostics and solutions (see here and here)

• A graphical dashboard for MRP

• Experimental design and time variation (see here)

• Instrumental variables with informative priors

• Regularizing using transformations and priors for a meta-analysis of survey incentives (see here)

• Risk aversion as backdoor Bayes

• Meta-analysis combining separate posterior distributions instead of raw estimates

• The gay penumbra: Measurement and time trends (see here)

• You need 16 times the sample size to estimate an interaction than to estimate a main effect (see here and here)

• Implicit assumptions in practically effective methods (see section 7.6 of BDA3)

• Conditional distributions and incoherent Gibbs (see here and here)

• Models for taxonomic structures as an example of the unfolding flower paradigm (see here)

• Domains of efficency for different methods of computing hierarchical linear and logistic regressions

• Effective number of parameters that can be estimated as a function of sample size, with phase transitions for some multilevel models

• Thinking about variation when hypothesizing a plausible average treatment effect (see here)

• A proposed schedule for post-publication review (see here)

• What is a replication?

• Design and sample size analysis for various identification strategies

• Two sides, no vig: The problem with generative and inferential reasoning in social science (see here)

• Using correlation to assess representativeness of a sample

• Displaying uncertainty and variation in hierarchical models

• The Squealer: A sensification of gradients in model fitting

• Phase space of computational algorithms for multilevel regression

• Priors on derived quantities

• Replacing hard constraints with soft constraints

• Advice about writing advice (see here)

• Scaling of group-level variances in hierarchical models (see here)

• Priors on derived quantities (see here)

• Iterative conditional imputation of missing data without nested loops

• Evidence-based practice is a two-way street (see here)

Back to Andrew Gelman's homepage.