---
title: "Multiblock process structure, profiles, and external validity"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Multiblock process structure, profiles, and external validity}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

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

## Conceptual feature blocks

```{r}
blocks <- process_feature_blocks(
  person_data,
  blocks = list(
    Psychometric = c("theta_hat", "mean_accuracy"),
    Gaze = c("mean_rt_ms", "mean_dwell_ms", "mean_revisits", "mean_aoi_entropy"),
    Pupil = c("mean_pupil_peak", "mean_pupil_auc"),
    Quality = c("mean_valid_gaze_prop", "mean_valid_pupil_prop")
  ),
  id = "person_id"
)

mfa <- fit_multiblock_process_map(blocks)
multiblock_person_coordinates(mfa)
multiblock_variable_coordinates(mfa)
multiblock_contributions(mfa)
plot(mfa, type = "individuals")
plot(mfa, type = "variables")
```

This is exploratory block-aware mapping, not a replacement for multimodal IRT.

## Process-profile discovery

```{r}
profiles <- fit_process_profile_mixture(
  person_data,
  variables = c("mean_dwell_ms", "mean_pupil_peak", "mean_aoi_entropy", "mean_revisits"),
  k = 3
)
process_profile_summary(profiles)
process_profile_probabilities(profiles)
plot(profiles, type = "profiles")
```

Profile names remain neutral until externally validated.

## External validity

```{r}
valid <- audit_process_external_validity(
  person_data,
  criterion = "independent_task_score",
  predictors = c("theta_hat", "mean_pupil_peak", "mean_dwell_ms", "mean_aoi_entropy"),
  baseline_predictors = c("age")
)

process_criterion_associations(valid)
incremental_process_validity(valid)
plot(valid, type = "associations")
plot(valid, type = "incremental")
```