---
title: "Calibration uncertainty and eye-tracking data quality"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Calibration uncertainty and eye-tracking data quality}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

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

Version 0.9 makes measurement quality visible in the analysis object. Calibration/validation error, successive-sample precision, effective sampling frequency, irregular sampling, and data loss can be summarized rather than hidden in preprocessing.

```{r, eval=FALSE}
cal <- read.csv(system.file("extdata","calibration_targets_demo.csv", package="eyeprocess"))
m <- calibration_error_model(cal)
gaze_uncertainty_ellipse(m)
plot(m)

g <- read.csv(system.file("extdata","gaze_quality_demo.csv", package="eyeprocess"))
q <- gaze_data_quality_profile(g, valid="valid", by="person_id")
data_quality_reporting_table(q)
```

`propagate_calibration_uncertainty()` and `probabilistic_aoi_assignment()` propagate empirical calibration error into AOI membership. These probabilities concern spatial membership under the error model; they are not probabilities of psychological attention.