Calibrate: does the test hold its alpha? | Catching a misspecified model | Recommend: which inference method? | Reporting | Putting it together: mp_plan()

Maximal models: correlated random slopes | Power from a fitted pilot model | Planning around a smallest effect size of interest | Safeguard power | Reading Type S and Type M errors

The three-function contract | Building a validated backend with mp_backend() | Using mp_scenario() and mp_power() | Built-in backends | Parallel grids | Further reading

Inference options: Wald vs LRT | Binary outcomes (binomial GLMM) | Sensitivity curve for binomial outcomes | Count outcomes (Poisson vs Negative Binomial) | Sensitivity curves (count outcomes)

Run analysis and capture manifest | Bundle results and export | Regenerating from manifest and seed | One-row manifest for saving

Glossary of main functions | Quick-start checklist | Wide input (strict and explicit) | Comparison with generic workflows

Prerequisites | Recipe 1: Random slope for a lagged predictor | Recipe 2: Bivariate lag sketch (two variables) | Recipe 3: Multivariate outcomes with mvbind (sketch) | See also

Contract: what tidyILD owns vs what it does not | Decision table | Handoff pattern: export after prepare, center, and lag | Code stubs (not evaluated) | dynamite (multivariate dynamic models) | PGEE (penalized GEE / high-dimensional longitudinal) | lavaan / blavaan (DSEM) | See also

Three axes before you fit anything | Decision flow (conceptual) | Feature map | Minimal examples | Further reading

Role of visual inspection | Map: question → function → bundle section (if any) | Example: spaghetti, predicted trajectories, facet by cluster | Recipe: facet panels without a dedicated helper | Partial effects for _wp and _bp (external packages) | See also

Estimands: population, partial pooling, and no pooling | ild_heterogeneity() | Diagnostics bundle and plots | Stratified descriptive comparison | See also

Why missingness matters in intensive longitudinal data | Types of missingness (useful labels) | Descriptive profiling: ild_missing_pattern() and heatmaps | Person-level compliance: ild_missing_compliance() | When to use ild_missing_model() and ild_missing_bias() | Complete-case vs mixed models (careful wording) | Cohort-level and hazard summaries | One entry point: ild_missingness_report() | MNAR as sensitivity (no single fix) | IPW and causal tools as one sensitivity route | Other templates (not evaluated here) | What tidyILD does not do (and where to look) | See also

Within-between decomposition | Irregular spacing and lags | Spacing classification

Decision axes | Comparison table | Short pointers to other vignettes | See also

Simulate and prepare | Inspect | Within-person centering and lags | Fit a model | Diagnostics and plots | MSM-style weights (IPTW and IPCW) — optional | Reproducibility

Simulation size and precision | What ild_simulate() encodes | Fixed-effect recovery with ild_power(..., return_sims = TRUE) | From recovery to power | Variance components (illustrative snapshot) | AR(1) in the DGP vs residual correlation in the fit | Bayesian and state-space extensions | Cross-backend validation harness (optional) | Limitations and scope | See also

When to use ild_ctsem() | Minimal workflow | Diagnostics and plots | Guardrails and reporting | Notes

Why this vignette exists | Identification assumptions | Estimand-first + history-builder workflow (v1) | Recovery harness | Inference caveats and strict mode | Notes on v1 scope

1. Fit model(s) | 2. Tidy fixed-effects table | 3. Fitted vs observed | 4. Residual diagnostics: ACF and Q-Q | With AR1 (nlme) | Time-varying effects (TVEM)

Ingesting a tsibble with ild_prepare() | What provenance is kept | How to inspect | Round-trip with ild_as_tsibble() | Limitations and policy

Motivation | What ild_kfas() assumes today | irregular_time and guardrails | Workflow sketch | Where to go next

What is a state-space model? | When use this instead of mixed-model residual correlation? | Filtered vs smoothed states | Minimal example | What the backend does not yet do | See also

See also

1. Prepare data | 2. Center and lag | 3. Fit model | 4. Run diagnostics | 5. Inspect ild_history() | 6. Generate ild_methods() | 7. Run ild_report() | 8. Export provenance | 9. Compare two pipelines