---
title: "SMLP 2026: Summer School Recap & Discussion"
subtitle: "10th Summer School on Statistical Methods for Linguistics and Psychology"
author:
  - name: "Job Schepens"
    affiliation:
      - name: Universität zu Köln
        department: CRC 1252 Prominence in Language
        city: Cologne, Germany
date: "16 September 2026"
date-format: "D MMMM YYYY"
format:
  revealjs:
    theme: [default, ../../stylesheets/mytheme.scss]
    logo: ../../logo.png
    slide-number: true
    scrollable: true
    mouse-wheel: true
    footer: "RDM Workshop Series · SFB 1252 Prominence in Language · Universität zu Köln"
    output-file: slides.html
    embed-resources: true
    transition: none
    background-transition: none
bibliography: "references.bib"
csl: "../24-latex-thesis-workflows/apa.csl"
---

<style>
@font-face {
  font-family: 'Albert Sans';
  font-style: normal;
  font-weight: 300;
  font-display: swap;
  src: url("../../stylesheets/fonts/AlbertSans-300.ttf") format("truetype");
}
@font-face {
  font-family: 'Albert Sans';
  font-style: normal;
  font-weight: 400;
  font-display: swap;
  src: url("../../stylesheets/fonts/AlbertSans-400.ttf") format("truetype");
}
@font-face {
  font-family: 'Albert Sans';
  font-style: normal;
  font-weight: 500;
  font-display: swap;
  src: url("../../stylesheets/fonts/AlbertSans-500.ttf") format("truetype");
}
@font-face {
  font-family: 'Albert Sans';
  font-style: normal;
  font-weight: 600;
  font-display: swap;
  src: url("../../stylesheets/fonts/AlbertSans-600.ttf") format("truetype");
}
@font-face {
  font-family: 'Albert Sans';
  font-style: normal;
  font-weight: 700;
  font-display: swap;
  src: url("../../stylesheets/fonts/AlbertSans-700.ttf") format("truetype");
}

/* Slide compacting tweaks to prevent vertical overflow */
.reveal table {
  font-size: 0.76em;
}
.reveal pre code {
  font-size: 0.72em;
  line-height: 1.25;
}
.reveal .callout {
  margin-top: 0.35em;
  margin-bottom: 0.35em;
  padding: 0.35em 0.7em;
  font-size: 0.85em;
}
.reveal h3 {
  margin-top: 0.15em;
  margin-bottom: 0.25em;
}
.reveal ul, .reveal ol {
  margin-top: 0.15em;
  margin-bottom: 0.15em;
}
.reveal li {
  margin-top: 0.1em;
  margin-bottom: 0.1em;
}
</style>

## Overview {.smaller}

::: {.columns}

::: {.column width="58%"}
### Session Agenda
- **SMLP** 
- **Participants** 
- **Keynotes** 
- **5 Classes** 
- **"Maximal Models"** 
- **Contrast Coding** 
- **CRC 1252** 
:::

::: {.column width="42%"}
:::

:::

---

## 1. SMLP {.smaller}

::: {.columns}

::: {.column width="50%"}
### 10-Years 
*   **Founded by** Shravan Vasishth (Uni Potsdam, 2017).
*   **Origin:** Funded as part of SFB 1287 (*Limits of Variability in Language*).
*   **Target audience:** cognitive scientists and linguists.
*   **Impact:** Over 1,000 participants over the years.
:::

::: {.column width="50%"}
### 2026 Edition
*   **5 parallel streams**.
*   **Keynotes:** Lisa DeBruine & Dale Barr.
*   **Many CRC 1252 Participants:** 
:::

:::

::: {.notes}
Speaker Notes:

:::

---


## Who Was Where? {.smaller}

::: {.columns}

::: {.column width="50%"}
### Which stream did you attend?
- [ ] Track 1: Intro to Bayesian (`brms`)
- [ ] Track 2: Advanced Bayesian (Stan)
- [ ] Track 3: Intro to R & Frequentist Stats
- [ ] Track 4: Foundational LMMs (`lme4`)
- [ ] Track 5: Advanced LMMs with Julia

### Keynotes attended?
- [ ] Lisa DeBruine (Power & Simulation)
- [ ] Dale Barr (PsyTeachR & Open Science)
:::

::: {.column width="50%"}
### Primary data in your CRC project:
*   *RTs* (eye-tracking, Self paced reading, etc.)
*   *Acoustic measurements* (f0, duration, intensity)
*   *Acceptability judgments* (binary, Likert, ordinal)
*   *Corpus frequencies & counts* (prominence markers)
*   *Electrophysiology* (EEG, ERPs, time-series)
:::

:::

::: {.notes}
Speaker Notes:
:::

---

## 2. Keynote 1: Lisa DeBruine {.smaller}

::: {.columns}

::: {.column width="52%"}
### Power and Simulation
*   **Analytic Power Fails:** Classical formulas (e.g. $G*\text{Power}$) assume independent data. They break down for **crossed random effects** (subjects $\times$ items) [@debruine2021understanding].
*   **Post-Hoc Power Fallacy:** Calculating power after the fact from observed effect size is just a noisy transformation of the $p$-value [@vasishth2018statistical].
*   **The Modern Solution:** Simulation-based power analysis.
:::

::: {.column width="48%"}
::: {.callout-tip}
### Core Principle
"If you can simulate a dataset that captures your design and hypothesized effect, you can calculate power for *any* model, no matter how complex."
:::

*   **Open Resources:**
    *   `faux` R package (factorial data simulation)
    *   [data-sim-workshops](https://debruine.github.io/data-sim-workshops/)
:::

:::

::: {.notes}
Speaker Notes:
- Emphasize Lisa DeBruine's point: stop using G*Power for psycholinguistic designs with both subject and item variance.
- Connect to CRC Workshop 18 on power analysis.
:::

---

## Simulation-Based Power Pipeline {.smaller}

::: {.columns}

::: {.column width="55%"}
```{mermaid}
%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '10px'}}}%%
flowchart TD
    A["Step 1: Specify Generative Model<br/>(Effects, SDs, Correlations)"] --> B["Step 2: Generate N Datasets<br/>(Synthetic Trials, Subjects, Items)"]
    B --> C["Step 3: Fit Model to Each Dataset<br/>(lme4 / brms / MixedModels)"]
    C --> D["Step 4: Evaluate Evidence Rate<br/>(Rate reaching p &lt; .05 or ROPE)"]
    D -->|Refine N / Design| A
```
:::

::: {.column width="45%"}
### Relevance for CRC 1252:
*   **Preregistration:** Justifying sample sizes rather than arbitrary $N=30$.
*   **Random effects:** Simulations reveal how item variance and random slopes diminish power for subtle prominence interactions.

::: {.callout-note}
### Previous CRC Workshop
Detailed in **[Workshop 18: Power Analysis](../18-power-analysis/)** & the interactive **[Sequential Simulation Report](https://jobschepens.github.io/samplesize/)**.
:::
:::

:::

::: {.notes}
Speaker Notes:
:::

---

## 3. Keynote 2: Dale Barr {.smaller}

::: {.columns}


:::

::: {.notes}
Speaker Notes:
- Mention PsyTeachR's free online textbooks.
:::

---

## 4. Track 1: Intro to Bayesian Analysis {.smaller}

::: {.columns}

::: {.column width="50%"}
### Instructors: Julia Haaf & Nicole Cruz
*   **Book:** *Bayesian Data Analysis for Cognitive Science* [@nicenboim2024introduction].
*   **Workflow:** Prior $\to$ Likelihood $\to$ Posterior.
*   **Syntax:** Familiar `lme4`-style syntax via `brms`:
    ```r
    brm(rt ~ prominence * focus + 
          (1 + prominence | subj) + 
          (1 + focus | item),
        data = dat, family = lognormal(),
        prior = priors)
    ```
:::

::: {.column width="50%"}
### Model Validation Workflow
1.  **Prior Checks:** Plausible RTs before fitting ([02_prior_checks](https://jobschepens.github.io/brms-ws/02_prior_predictive_checks_rt.html)).
2.  **Diagnostics:** $\hat{R} \le 1.01$, $\text{ESS} > 400$ ([08_convergence](https://jobschepens.github.io/brms-ws/08_convergence.html)).
3.  **Posterior Checks:** Test fit with `pp_check` ([03_posterior_checks](https://jobschepens.github.io/brms-ws/03_posterior_predictive_checks_rt.html)).
4.  **Inference:** Credible intervals and ROPE ([06_rope](https://jobschepens.github.io/brms-ws/06_rope.html) · [Workshop 14](../14-bayesian-rope-testing/)).

:::

:::

::: {.notes}
Speaker Notes:
- Explain Track 1: Ideal entry point for researchers moving from lme4 to brms.
- Point out that lognormal distribution natively accounts for RT right-skew.
:::

---

## Track 1: Regularizing Priors {.smaller}

::: {.columns}

::: {.column width="50%"}
### Prior Specification in `brms`
```r
priors <- c(
  # Baseline intercept ~ 400 ms
  prior(normal(6, 0.5), class = "Intercept"),
  
  # Plausible effect size bounds
  prior(normal(0, 0.1), class = "b"),
  
  # Regularize random slope variances
  prior(exponential(2), class = "sd"),
  
  # Discourage extreme +/- 1 correlations
  prior(lkj(2), class = "cor")
)
```
:::

::: {.column width="50%"}
### Why Priors help
*   **Singularity in lme4:** estimated correlations between intercepts and slopes often get to $\pm 1.00$.
*   **The LKJ Prior:** Setting $\text{LKJ}(\eta = 2)$ concentrates probability mass around 0, shrinking extreme correlations and stabilizing estimation [@nicenboim2024introduction].

::: {.callout-note}
### Prior Sensitivity Tutorial
Compare flat vs. regularizing priors in **[brms-ws 04_comparing_priors](https://jobschepens.github.io/brms-ws/04_comparing_priors_rt.html)**.
:::
:::

:::

::: {.notes}
Speaker Notes:
- The LKJ prior is one of the most practical Bayesian solutions to frequentist convergence errors.
:::

---

## 5. Track 2: Advanced Bayesian Modeling {.smaller}

::: {.columns}

::: {.column width="50%"}
### Instructors: Bruno Nicenboim & Johan Hennert
*   **Beyond `brms`:** Writing custom probabilistic programs in native Stan.
*   **Cognitive Process Models:**
    *   *Drift-Diffusion Models (DDM):* Jointly model decision bounds, drift rate, and non-decision time.
    *   *Log-Normal Race Models:* Competing accumulators for lexical or syntactic selection.
:::

::: {.column width="50%"}
### Principled Model Comparison
*   **Bayes Factors (Bridge Sampling):**
    *   Tests point nulls ($H_0: \theta = 0$). Prior sensitive!
    *   See **[Workshop 15: Bayes Factors](../15-bayes-factor/)** & **[07_bayes_factors](https://jobschepens.github.io/brms-ws/07_bayes_factors.html)**.
*   **PSIS-LOO-CV (`loo` package):**
    *   Out-of-sample predictive accuracy.
    *   See **[Workshop 13: Cross-Validation](../13-bayesian-model-comparison-cv/)** & **[05_loo](https://jobschepens.github.io/brms-ws/05_loo.html)**.

::: {.callout-note}
### Decision Frameworks Comparison
Explore **[ROPE vs. Bayes Factor vs. LOO](https://jobschepens.github.io/brms-ws/06_sequential_testing.html)**.
:::
:::

:::

::: {.notes}
Speaker Notes:
- Distinguish Bayes Factors (testing hypotheses) from LOO-CV (predictive generalization).
:::

---

## 6. Track 4: Frequentist Foundations & LMMs {.smaller}

::: {.columns}

::: {.column width="50%"}
### "Keep It Maximal" [@barr2013random]
*   Include all random intercepts and slopes for within-unit experimental factors.
*   **Goal:** Rigorously protect against inflated Type I error ($\alpha > .05$).
*   **Trade-off:** Frequently results in overparameterization, singular fits, and optimizer convergence errors.
:::

::: {.column width="50%"}
### "Parsimonious Models" [@matuschek2017balancing]
*   Select random effects via backward elimination with Likelihood Ratio Tests (LRT).
*   **Finding:** Over-parameterized maximal models cause **substantial loss of power**.
*   **Recommendation:** Keep only random slopes supported by the empirical data.
:::

:::

::: {.notes}
Speaker Notes:
-  Barr et al. maximal vs. Matuschek et al. parsimonious.
- Takeaway: Forcing unsupported slopes destroys power for the fixed effect you actually care about.
:::

---

## Contrast Coding {.smaller}

::: {.columns}

::: {.column width="50%"}
### R Default (Treatment Coding)
```r
# Default R: contr.treatment
fit <- lmer(RT ~ Prominence * Focus + 
              (1 | Subj), data = dat)
```
*   `Prominence` coefficient is **NOT** the main effect of prominence; it is the difference between the reference level and the other level.
*   Interactions make the interpretation of lower-order terms more complex [@schad2020capitalize].
:::

::: {.column width="50%"}
### Sum Coding
```r
# Sum contrasts
contrasts(dat$Prominence) <- contr.sum(2)
contrasts(dat$Focus) <- contr.sum(2)
```
*   Intercept ($\beta_0$) = Grand mean.
*   Slopes ($\beta_1, \beta_2$) = True main effects averaged across all levels of the other factors.

:::

:::

::: {.notes}
Speaker Notes:
- Schad et al. (2020) 
- Never report "main effects" from models with interactions without verifying contrast coding!
:::

---

## Contrast Coding Overview {.smaller}

| Contrast Scheme | R Command | Intercept ($\beta_0$) | Slope ($\beta_1, \beta_2, \dots$) |
|---|---|---|---|
| **Treatment (Default)** | `contr.treatment(k)` | Reference group mean | Difference: Group $i$ minus Reference |
| **Sum** | `contr.sum(k)` | Grand mean across groups | Deviation: Group $i$ minus Grand mean |
| **Successive Difference** | `MASS::contr.sdif(k)` | Grand mean across groups | Difference between adjacent successive levels ($\mu_i - \mu_{i-1}$) |
| **Helmert** | `contr.helmert(k)` | Grand mean across groups | Level $i$ compared to mean of prior levels |

::: {.callout-important}
### Good Practice
Check `contrasts(dat$my_factor)` before fitting models to check coding.
:::

::: {.notes}
Speaker Notes:
- Reiterate that changing contrast coding changes parameter estimates, SEs, and p-values of lower-order terms.
- Successive difference coding (`MASS::contr.sdif`) is ideal for ordered psycholinguistic data (e.g., word-by-word reading times, spillover regions, developmental stages) to test step-wise transitions between adjacent conditions.
:::

---

## 7. Track 5: Advanced LMMs with Julia {.smaller}

::: {.columns}

::: {.column width="50%"}
### Instructors: Reinhold Kliegl & Phillip Alday
*   **The Engine:** `MixedModels.jl` [@kliegl2024embrace].
*   **Why Julia?**
    *   No two-language overhead (pure Julia).
    *   High-performance sparse matrix algebra.
    *   **10x to 100x faster** than `lme4`.
*   **Ideal For:** Massive linguistic corpora and multi-million token datasets.
:::

::: {.column width="50%"}
### Standout Capabilities
*   **Random Effects PCA (`rePCA`):** Diagnoses the exact mathematical dimensionality of variance components, exposing zero-variance slopes.
*   **Fast Parametric Bootstrap:** Generates 10,000 bootstrap replicates in minutes for accurate non-asymptotic CIs and $p$-values.
:::

:::

::: {.notes}
Speaker Notes:
- Explain that MixedModels.jl is the state of the art for frequentist mixed models at scale.
:::

---

## Interoperability: Calling Julia from R {.smaller}

::: {.columns}

::: {.column width="55%"}
```r
library(JuliaCall)
julia_setup()
julia_library("MixedModels")
julia_library("DataFrames")

# Transfer R data frame to Julia in-memory
julia_assign("df", crc_data)

# Fit model at compiled C-speed
julia_command("m1 = fit(MixedModel, 
  @formula(RT ~ Prom * Ctx + (1 + Prom | Subj)), 
  df)")

# 1,000 bootstraps in seconds
boot <- julia_eval("parametricbootstrap(1000, m1)")
```
:::

::: {.column width="45%"}
::: {.callout-tip}
### Seamless R/Julia Workflow
1.  Wrangle & clean in R (`tidyverse`)
2.  Fit & bootstrap in Julia (`MixedModels.jl`)
3.  Plot & report in R (`ggplot2`)
:::
*   No need to rewrite your entire project workflow in Julia!
:::

:::

::: {.notes}
Speaker Notes:
- JuliaCall allows researchers to keep their R tidyverse workflow while using Julia purely as a compute engine.
:::

---

## 8. Tool Comparison? {.smaller}

| Feature | R: `lme4` | R: `brms` (Stan) | Julia: `MixedModels.jl` |
|---|---|---|---|
| **Paradigm** | Frequentist (REML/ML) | Bayesian (MCMC/NUTS) | Frequentist (REML/ML) |
| **Speed** | Moderate | Slow (minutes to hours) | Fast (seconds) |
| **Singular Fits** | Frequent warning | Regularized by priors | Diagnosed via `rePCA` |
| **Distributions** | Gaussian, Binomial, Poisson | Any | Continuous, Binomial, Poisson |
| **Inference** | Wald, LRT, Kenward-Roger | Credible intervals, ROPE, BF | Parametric Bootstrap |
| **Setup Cost** | Standard R | Requires C++ toolchain | Requires Julia install |

::: {.notes}
Speaker Notes:
:::

---

## Resolving Singular Fits in Practice {.smaller}

::: {.columns}

::: {.column width="52%"}
::: {.column width="58%"}
```{mermaid}
%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '10px'}}}%%
flowchart TD
    A["Singular Fit Warning in lme4"] --> B{"Overparameterized<br/>correlation?"}
%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '13px', 'fontFamily': 'Albert Sans, sans-serif'}}}%%
flowchart LR
    A["Singular Fit Warning<br/>in lme4"] --> B{"Overparameterized<br/>correlation?"}
    B -->|Yes| C["Zero-Correlation Model:<br/>(1 + X || Group) in lme4"]
    B -->|Yes| D["Bayesian Prior:<br/>prior(lkj(2), class='cor') in brms"]
    A --> E{"Unsupported<br/>random slopes?"}
    E -->|Yes| F["Parsimonious Model:<br/>Backward LRT selection"]
    E -->|Inspect| G["Julia rePCA():<br/>Diagnose zero-variance axes"]
```
:::

::: {.column width="48%"}
::: {.column width="42%"}
### Practical Decision Guide
1.  **Don't ignore it:** A singular fit means estimates are on the boundary of parameter space.
2.  **Correlation near $\pm 1$:** Use `||` syntax in `lme4` or `lkj(2)` in `brms`.
3.  **Variance near 0:** Drop unsupported slopes to recover power.

::: {.callout-note}
### Testing Null Effects
For determining if an effect is practically absent, see **[Workshop 14: ROPE Testing](../14-bayesian-rope-testing/)** & **[06_rope](https://jobschepens.github.io/brms-ws/06_rope.html)**.
:::
:::

:::

::: {.notes}
Speaker Notes:
- Remind attendees that singular fits are not fatal, but signal that data cannot support all parameters.
:::

---

## Shared Bayesian Core: SMLP Book vs. RDM Workshops {.smaller}

The foundational `brms` methods taught at SMLP directly mirror our SFB 1252 RDM series:

| Core Bayesian Topic | SMLP Book Chapters (Nicenboim et al.) | Relevant SFB 1252 RDM Workshops & Notebooks |
|---|---|---|
| **Bayesian Workflow & Priors** | [Ch. 4: Regression Models](https://bruno.nicenboim.me/bayescogsci/ch-reg.html#sec-brms) · [Ch. 5: Hierarchical](https://bruno.nicenboim.me/bayescogsci/ch-hierarchical.html#sec-brms-hierarchical) | **[Workshop 11: Regression with brms](../11-bayesian-regression-models/)** · [01_setting_priors](https://jobschepens.github.io/brms-ws/01_setting_priors.html) |
| **Prior & Posterior Checks** | [Ch. 4: Prior Checks (§4.2)](https://bruno.nicenboim.me/bayescogsci/ch-reg.html#sec-priorpred) · [Ch. 4: Posterior Checks (§4.4)](https://bruno.nicenboim.me/bayescogsci/ch-reg.html#sec-ppchecks) | **[Workshop 11](../11-bayesian-regression-models/)** · [02_prior_checks](https://jobschepens.github.io/brms-ws/02_prior_predictive_checks_rt.html) · [03_posterior_checks](https://jobschepens.github.io/brms-ws/03_posterior_predictive_checks_rt.html) |
| **MCMC Convergence & Diagnostics** | [Ch. 3: Computational BDA (§3.3)](https://bruno.nicenboim.me/bayescogsci/ch-compbda.html#sec-diagnostics) ($\hat{R}$, ESS, divergences) | **[Workshop 11](../11-bayesian-regression-models/)** · [08_convergence](https://jobschepens.github.io/brms-ws/08_convergence.html) |
| **Model Comparison (LOO-CV)** | [Ch. 12: Comparison](https://bruno.nicenboim.me/bayescogsci/ch-comparison.html) · [Ch. 14: Cross-Validation](https://bruno.nicenboim.me/bayescogsci/ch-cv.html#sec-cvbrms) | **[Workshop 13: Cross-Validation](../13-bayesian-model-comparison-cv/)** · [05_loo](https://jobschepens.github.io/brms-ws/05_loo.html) |
| **Bayes Factors & Point Hypotheses** | [Ch. 13: Bayes Factors (§13.1)](https://bruno.nicenboim.me/bayescogsci/ch-bf.html#sec-bfsavage) (Bridge sampling / Savage-Dickey) | **[Workshop 15: Bayes Factors](../15-bayes-factor/)** · [07_bayes_factors](https://jobschepens.github.io/brms-ws/07_bayes_factors.html) |
| **Simulation-Based Power** | [Ch. 3: Power & Sample Size (§3.5)](https://bruno.nicenboim.me/bayescogsci/ch-compbda.html) (Generative models) | **[Workshop 18: Power Analysis](../18-power-analysis/)** · [Sequential Simulation Report](https://jobschepens.github.io/samplesize/) |


::: {.notes}
Speaker Notes:

:::

---

## SMLP Bayesian Topics Not in RDM Yet {.smaller}

SMLP covers advanced response families, custom syntax, and cognitive process models:

::: {.columns}

::: {.column width="50%"}
### Advanced `brms` Distributions & Syntax
*   **Shifted Lognormal & Ex-Gaussian RTs:**
    *   Estimate non-decision time ($ndt$) alongside mean/variance: `family = shifted_lognormal()` ([Book Ch. 7](https://bruno.nicenboim.me/bayescogsci/ch-reg.html)).
*   **Ordinal & Cumulative Models:**
    *   Directly model Likert / acceptability scales: `family = cumulative("logit")`.
*   **Non-Linear Syntax (`bf()`):**
    *   Fit non-linear curves & saturation functions.
*   **Missing Data Imputation:**
    *   Model missing predictors via `brm(y ~ mi(x) + ...)`.
:::

::: {.column width="50%"}
### From `brms` to Stan & Cognitive Models
*   **Inspecting Generated Stan Code:**
    *   `make_stancode()` and `make_standata()` to inspect what `brms` compiles ([Book Ch. 8](https://bruno.nicenboim.me/bayescogsci/ch-introstan.html)).
*   **Custom Families (`custom_family()`):**
    *   Define user-specified likelihoods in `brms`.
*   **Computational Cognitive Models (Track 2):**
    *   *Drift-Diffusion Models (DDM):* Joint choice & RT.
    *   *Race Models:* Competing accumulators ([Book Ch. 18](https://bruno.nicenboim.me/bayescogsci/ch-lognormalrace.html)).
    *   *Mixture Models:* Latent subpopulations ([Book Ch. 17](https://bruno.nicenboim.me/bayescogsci/ch-mixture.html)).
:::

:::

::: {.callout-note}
### Future Bayesian Topics for Project S
:::

::: {.notes}
Speaker Notes:
- Focus strictly on the Bayesian modeling gaps.
- Shifted lognormal is especially useful in linguistics because RTs have an absolute biological lower bound (~150-200ms non-decision time).
- Mention that inspecting Stan code via make_stancode() is a great bridge from brms to native Stan.
:::

---

## RDM Bayesian Topics Not Covered in SMLP {.smaller}

Our local RDM series emphasizes practical equivalence testing, sequential designs, and post-processing tools:

::: {.columns}

::: {.column width="50%"}
### The ROPE Decision Framework
*   **Practical Equivalence Testing:**
    *   Evaluate whether effects are practically zero, negligible, or present (**[Workshop 14: ROPE Testing](../14-bayesian-rope-testing/)**).
*   **HDI + ROPE Decision Rules:**
    *   Three-way verdicts (PRESENT, ABSENT, UNDECIDED) with `bayestestR` ([06_rope](https://jobschepens.github.io/brms-ws/06_rope.html)).
*   **Defensible Boundaries:**
    *   Setting domain-justified equivalence limits in linguistics rather than relying only on point nulls.
:::

::: {.column width="50%"}
### Sequential Designs & Post-Processing
*   **Benchmarking Decision Frameworks:**
    *   Direct side-by-side comparison of **ROPE vs. Bayes Factor vs. LOO-CV** in sequential testing (**[Workshop 15](../15-bayes-factor/)**, [06_sequential_testing](https://jobschepens.github.io/brms-ws/06_sequential_testing.html)).
*   **Post-Hoc Contrasts with `emmeans`:**
    *   Pairwise comparisons and marginal predictions on `brms` fits with `emmeans` and `marginaleffects`.
*   **Containerized `brms` Workflows:**
    *   Ready-to-run Docker & Ramses HPC setups in **[brms-ws](https://github.com/jobschepens/brms-ws)** to bypass local C++ compiler setup.
:::

:::

::: {.callout-tip}
### Complementary Bayesian Perspectives
SMLP focuses on Bayes Factors; RDM focuses on practical equivalence (ROPE)
:::

::: {.notes}
Speaker Notes:
:::

---

## References {.smaller}

::: {#refs}
:::