Gazepoint AOI Sensitivity Analysis Plan

```{r setup, include=FALSE} knitr::opts_chunk$set(collapse = TRUE, comment = “#>”, eval = FALSE) library(gp3tools) ```

Purpose

This template fixes the Gazepoint-to-core conversion and AOI sensitivity choices before branch-specific results are inspected. `gp3tools` remains a thin adapter: the scientific perturbation, assignment, feature, model, and reporting logic is delegated to `eyeprocess`.

Pre-analysis decisions

Record:

• the exact Gazepoint x/y source columns;
• source coordinate unit (`normalized` or `px`);
• screen width/height used for normalized-to-pixel conversion;
• physical screen size and viewing distance for degree-based perturbations;
• sample versus fixation observation level;
• nominal AOI source/version;
• perturbation grid and units;
• overlap and screen-boundary policies;
• fixed downstream model callback;
• failure/non-convergence handling;
• planned assignment, coefficient, interval, convergence, and model-N summaries.

Filled synthetic plan

```yaml analysis_id: gazepoint-disclosure-aoi-robustness source_columns: x: FPOGX y: FPOGY coordinate_unit: normalized screen: width_px: 1920 height_px: 1080 physical_width: 53.1 physical_height: 29.9 viewing_distance: 60 observation_level: sample perturbations: unit: deg dilation: [0.25, 0.50, 1.00] erosion: [0.25] translation_x: [0.50] translation_y: [0.50] overlap_policy: ambiguous model: fixed_callback: true failure_handling: keep_geometry_failures: true keep_model_failures: true keep_nonconverged_rows: true ```

Adapter call

```{r planned-run} result <- run_gazepoint_aoi_sensitivity( gaze, aois, gaze_x_col = “FPOGX”, gaze_y_col = “FPOGY”, coordinate_unit = “normalized”, perturbation_unit = “deg”, dilations = c(.25, .50, 1.00), erosions = .25, translations_x = .50, translations_y = .50, screen_width_px = 1920, screen_height_px = 1080, viewing_distance = 60, physical_screen_size = c(53.1, 29.9), observation_id_col = “sample_id”, participant_col = “participant”, trial_col = “trial”, duration_col = “duration”, observation_level = “sample”, overlap_policy = “ambiguous” ) ```

Interpretation rule

A high unchanged-assignment proportion or stable coefficient describes robustness only to the declared Gazepoint coordinate conversion and AOI perturbation plan. It does not establish that the nominal AOI boundaries are correct.

At sample level, zero means valid gaze samples existed but none entered the AOI. Missing means there was no valid assignment opportunity. Do not convert one into the other.

Reporting example

Gazepoint normalized coordinates were converted using the prespecified screen dimensions, and AOI sensitivity was evaluated at the sample level using a fixed degree-based perturbation plan. We report assignment stability, coefficient ranges and confidence intervals, model convergence and N variation, and all failed geometry/model branches. The adapter delegated scientific calculations to `eyeprocess`.

This is a synthetic reporting template, not an empirical finding.

API map

• `audit_gazepoint_aoi_uncertainty()` — validate columns, units, and adapter boundary;
• `run_gazepoint_aoi_sensitivity()` — delegate the complete sensitivity workflow;
• `plot_gazepoint_aoi_sensitivity()` — delegated geometry, assignment, coefficient, and surface diagnostics.

Use the main Gazepoint AOI Perturbation and Uncertainty article for troubleshooting, interpretation, and limitations.