--- title: "Visual-context and testlet IRT" output: rmarkdown::html_vignette vignette: > %\VignetteIndexEntry{Visual-context and testlet IRT} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r setup, include=FALSE} knitr::opts_chunk$set(collapse = TRUE, comment = "#>", eval = FALSE) library(eyeprocess) ``` Items displayed on the same screen, page, diagram, or stimulus can share presentation-context variance. `fit_visual_context_irt()` models that known design dependence as a separate testlet/context dimension rather than silently treating every item as independent. ```{r} registry <- visual_context_registry( item_metadata, item = "item_id", context = "screen_id" ) fit <- fit_visual_context_irt( response_matrix, registry = registry, context = "screen_3" ) compare_visual_context_irt(fit) context_factor_effects(fit) audit_visual_context_dependence(fit) plot(fit, type = "loadings") plot(fit, type = "difficulty_change") plot(fit, type = "context_registry") ``` The context dimension represents shared presentation context unless independent substantive theory and validation justify another interpretation.