Package: MAIHDA 0.3.0

Introduction | Installation | The data | The maihda() workflow | Visual diagnostics | A crossed-dimensions alternative | A contextual cross-classified model | Design-weighted MAIHDA (survey data) | Comparing across groups | Under the hood: the building blocks | Fit a single model | A custom adjusted model and the PCV | Stepwise PCV | Order-invariant attribution: Shapley values and dominance analysis | Discriminatory accuracy and the response-scale VPC | The group comparison directly | Where to next | Interactive Shiny App | References

Sparse intersectional cells | A dataset with a known interaction | What lme4 reports | What brms adds: posterior uncertainty | Comparison | Why the interval, not the point, is what changes

From a snapshot to a trajectory | The data | Fitting and the time-varying VPC | Two VPCs, and which one to report | Decomposing the trajectory: additive vs. multiplicative | Scope and cautions | Reference

Before you fit | Is MAIHDA the right tool? | The central tradeoff: more dimensions means emptier cells | One call: maihda_describe() | What sparse cells do: singular fits | Continuous variables and the analytic sample | What the summaries can and cannot tell you | Which engine, which design? | A suggested learning path | References

Two different things called "cross-classified" | Part 1, The crossed-dimensions decomposition | Running a crossed-dimensions analysis | Figures | Comparing across a higher-level group | A Bayesian fit | Two important caveats | Part 2, Contextual cross-classified MAIHDA (context =) | With the full maihda() workflow | context = vs. group = | Notes

Why binary outcomes? | Fitting a logistic MAIHDA model | The VPC is on the latent scale | Adjusted model and PCV | What "interaction" means on the logit scale | Discriminatory accuracy (AUC and Median Odds Ratio) | Plots adapt to the binomial family | Count outcomes work the same way | References

Introduction | Example data: countries, gender, and socioeconomic status | One-call workflow | Visualizing the comparison | Share versus magnitude | Additive share (PCV) by group | Direct group comparison | Adding bootstrap intervals

From a fitted model to a manuscript | glance() -- the one-row headline | tidy() -- estimates as a tidy tibble | maihda_table() -- the two canonical write-up tables | Choosing a model structure with maihda_ic() | See also | References

Overview | Run a standard analysis and choose the multiplicity rule | Highlight flagged strata | Is an interaction negligible? (equivalence / ROPE) | See also | References

Overview | vpc -- variance partition | predicted -- stratum predictions with intervals | obs_vs_shrunken -- shrinkage made visible | effect_decomp -- additive vs. intersection-specific | prediction_deviation -- the deviation dashboard | Group-comparison plots | Customizing the appearance | See also

Data preparation | Recode and collapse strata | Fit the MAIHDA | Which strata drive the pattern? | How many interactions actually matter? A ROPE | Plot the case study | Results | References

Introduction | Launching the Application | Online Version | Local Version | App Features | 1. Data Import | 2. Variable Selection & Strata Creation | 3. Model Fitting & Settings | 4. Interactive Visualizations | 5. Stepwise Variance Analysis (PCV) | Summary