---
title: "Specifying and Fitting the M4 Reference Model"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Specifying and Fitting the M4 Reference Model}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
library(eyeprocess)
```

The reference model uses a marginalized finite-state Markov process over explicitly ordered participant/session sequences. K=1 is a formal null and K=2 is the conservative default.

```{r}
sim <- simulate_multimodal_m4(n_person = 30, n_item = 8, seed = 20260820)
audit <- audit_multimodal_m4_identifiability(sim, include_posterior = FALSE)
audit
```

A deliberate CmdStanR fit is required for posterior inference; documentation does not compile or sample automatically.

```{r eval=FALSE}
fit <- fit_multimodal_m4(sim, spec = multimodal_m4_spec())
summary(fit)
fitted(fit, type = "state")
```