---
title: "Complete Pupillometry Pipeline Walkthrough"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Complete Pupillometry Pipeline Walkthrough}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

<style>
img:not(.logo) {
  background-color: white;
}
</style>

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 7,
  fig.height = 5,
  fig.dpi = 300
)
```

## 📦 Introduction

This vignette will walk you through the basics of getting started with the
`eyeris` package to preprocess your (eyelink) pupillometry data from start to finish.
Here, we'll walk through loading in the raw data, preprocessing, and quality
control in a step-by-step manner. 

We will specifically focus on demonstrating the
`eyeris::glassbox()` function, which we built to streamline, in part, the process
of collecting and preprocessing any freshly collected **raw** pupillometry dataset
within minutes to assess the quality of those data while simultaneously running
a full preprocessing pipeline in _1/2-ish_ lines of code.

## 🔎 The Glass Box Function

Our `glassbox` function -- in contrast to a "black box" function where you run it
and obtain a result with little-to-no idea as to how you got from raw input to
cleaned output -- provides a highly opinionated prescription of steps and starting
parameters we (computational neuroscientists and signal processing experts at 
the Stanford Memory Lab) strongly believe any pupillometry researcher should use
as their default starting point when preprocessing any set of pupillometry data.

**We recommend starting out with the `glassbox` function to (ideally) reduce the**
**probability of accidental mishaps when "reimplementing" the individual steps**
**that make up the preprocessing pipeline both within _and_ across projects.**

Moreover, the `glassbox` function provides an "interactive" framework where you 
can evaluate the consequences of the parameters used within each step on your data in real time, therefore facilitating a fairly easy-to-use workflow for parameter optimization
on your particular dataset. 

In essence, this _interactive_ process takes each
opinionated step and allows you to see both pre and post timeseries plots so you
can adjust parameters stepwise until you are satisfied with your choices (and
their consequences on your pupil timeseries data).

### Installing `eyeris`

If you haven't already installed the `eyeris` package:

<!-- # nolint start -->
```{r setup, eval=FALSE}
# Install from CRAN (coming soon)
# install.packages("eyeris")

# pak
# install.packages("pak")
# pak::pak("shawntz/eyeris")

# Or install the development version from GitHub
# install.packages("devtools")
# devtools::install_github("shawntz/eyeris")
```

```{r pkgs, echo=FALSE, message=FALSE, warning=FALSE}
# install.packages("devtools")
# devtools::install_github("shawntz/eyeris")
```
<!-- # nolint end -->

### Loading `eyeris` Package

```{r}
library(eyeris)
```


### Loading Your Raw Data

For this demo, we'll use our built in demo dataset, which contains a handful of
trials from an associative memory task recorded in our lab.

```{r load-data}
demo_data <- system.file("extdata", "memory.asc", package = "eyeris")
```

### Running the Fully-Automated Pipeline

Here, we use the example data along with the default prescribed parameters 
and pipeline recipe:

```{r out.width='100%'}
# Run an automated pipeline with no real-time inspection of parameters
output <- eyeris::glassbox(demo_data)

# Preview first and last steps of the pipeline
plot(
  output,
  steps = c(1, 5),
  preview_window = c(0, nrow(output$timeseries$block_1)),
  seed = 0
)
```

### Running the Pipeline Interactively

```{r eval=FALSE, out.width='100%'}
output <- eyeris::glassbox(demo_data, confirm = TRUE, seed = 0)
```

### Overriding the Default Parameters

To override the default `glassbox` parameters directly within the `glassbox()`
function call, you need to pass in the appropriate parameter(s) for each pipeline
step to `glassbox()`.

#### Example

```{r}
output <- eyeris::glassbox(
  demo_data,
  confirm = FALSE, # TRUE if you want to visualize each step in real-time
  deblink = list(extend = 40),
  lpfilt = list(plot_freqz = FALSE)
)
```

##### Pipeline Steps with Overridable Parameters

1. `eyeris::load_asc()`: `block`
2. `eyeris::deblink()`: `extend`
3. `eyeris::detransient()`: `n`; `mad_thresh`
4. `eyeris::lpfilt()`: `wp`; `ws`; `rp`; `rs`; `plot_freqz`

## 💬 Caveats

### Detrending Step

`detrend_data` is set to `FALSE` by default. Detrending your pupil timeseries can 
have unintended consequences; we thus recommend that users understand the 
implications of detrending – in addition to whether detrending is appropriate 
for the research design and question(s) – before using this function.

### _I received this message... what does it mean and what should I do?_

You (hopefully) shouldn't encounter this, but if you do, let us explain:

```
***WARNING: SOMETHING OUTRAGEOUS IS HAPPENING WITH YOUR PUPIL DATA!*** 

The median absolute deviation (MAD) of your pupil speed is 0.
 This indicates a potential setup / hardware issue which has likely propogated into your pupil recording.
 Most often, this is due to online filtering being applied directly to your pupil data via the EyeLink Host PC.

 WE DO NOT RECOMMEND USING ANY ONLINE AUTO FILTERING DURING YOUR DATA RECORDING.

 ***ADDITIONAL INFO***
The default setting for communicating with the EyeLink
 machine is to have filtering enabled on either live-streamed data or saved data in EDF files.

 We do not want these filters because:
 (1) it is unclear how the EyeLink machine is specifying their low-pass filter, and
 (2) it conflicts with an assumption here in `eyeris::detransient()` to use one-step 
difference to compute a threshold for detecting large jumps in pupil data
 (suggesting blinking or other artifacts).

 ***GENERAL TIP***
 You should aim to have your data as raw as possible so you can 'cook' it fresh!

 ***WHAT TO DO NEXT***
 If you would like to continue preprocessing your current pupil data file,
 you should consider skipping the `eyeris::detransient()` step of the pipeline altogether.
 Advanced users might consider additional testing by manually passing in different `mad_thresh` override values
 into `eyeris::detransient()` and then carefully assessing the consequences of any given override value on the pupil data;
 however, this step is not recommended for most users.

 PLEASE CONTINUE AT YOUR OWN RISK.
```

Basically the default setting for communicating with an EyeLink machine is to 
have filtering enabled on either live-streamed data or saved data in EDF files. 
You **do not** want these filters because 

  1. it’s unclear what are their low-pass filter specifications, and
  2. it conflicts with an assumption in `eyeris::detransient()` to use one-step 
  difference to compute a threshold for detecting large jumps in pupil data 
  (suggesting blinking or other artifacts). 
  
> In general, you’d like to have your data as raw as possible so you can "cook" it fresh.

**So, if your data aren't as raw as possible, you will need to run `eyeris` without the `detransient` step.**

To skip the `detransient` step, pass `skip_detransient = TRUE` into your 
`glassbox()` call.

#### Some additional notes about preventing live filtering in future recordings...

Unfortunately the `Live Sample Filter` and `Saved Sample Filter` are two common 
settings when experiment softwares communicate with EyeLink. For example, 
`PsychoPy` incorporates them in the eyetracking setting panel, and if you use 
`pylink` directly, these two settings are also included in their example script / 
function used to connect to EyeLink. The same goes for `Eprime` and `NBS Presentation`, 
where these two filter settings are users’ responsibilities to set. 

Adjusting the `config.ini` file on your EyeLink host PC to set the defaults to be `OFF` 
is not sufficient as they will get overwritten when experiment softwares 
(`PsychoPy`, `Eprime`, `NBS Presentation`, etc.) make the initial connect request. 

Therefore, unfortunately all users for all future experiments need to be cognizant 
of this setting. This is why `eyeris::detransient()` will yell the long warning message
at you (as a reminder to turn off hardware filters).

So moving forward, please consider taking the following recommendations into consideration:

- Use `pylink` directly, add additional optional settings beyond what the SR Research 
example scripts set so you ping down all the optional settings. In terms of the 
filter settings, it boils down to calling `pylink.EyeLink.setHeuristicLinkAndFileFilter()`, 
but there are other settings that you likely want to hard-code into your experiments.

- Use `pylink` directly, but don’t code additional optional settings into you scripts. 
Instead, you must manually inspect the settings on the console screen of host PC 
**at the start of each experiment** to make sure they are set correctly. Specifically, 
in terms of the filter settings, you just need to make sure both `Online` and 
`Saved Sample Filtering` are set to `OFF` on the EyeLink setup screen. 
_It’s on the left column along with the button to auto-threshold and toggle which eye to track._ 
**This applies to all the other optional settings too.**

- Use the `builder` and `ioHub` directly, and set these settings in the eyetracking 
setting panel, which saves to each experiment, so you don’t need to worry about it. 
Specifically, in terms of the filter settings, you should set `FILTER_LEVEL_OFF` 
for both **live** and **saved sample filtering**. You can see all the different
optional settings here: [https://github.com/mh105/psychopy-eyetracker-sr-research/blob/main/psychopy_eyetracker_sr_research/sr_research/eyelink/eyetracker.py](https://github.com/mh105/psychopy-eyetracker-sr-research/blob/main/psychopy_eyetracker_sr_research/sr_research/eyelink/eyetracker.py).

- **For most users**, we simply recommend using the `PsychoPy` builder for 
`PsychoPy` experiments and use the `psychopy-eyetracker-sr-research` plug-in
maintained by the `PsychoPy` development team + [Alex He, Ph.D.](https://github.com/mh105), Stanford Postdoctoral Scholar in the Purdon lab under the Department of Anesthesiology 
and in the Wagner lab under the Department of Psychology, and 
_co-author of_ `eyeris`.

```{r, echo=FALSE, out.width='80%', fig.align='center'}
"psychopy-eyetracker-sr-research-plug-in.png" |>
  knitr::include_graphics()
```

---

## 📚 Citing `eyeris`

<div class="alert alert-light" style="margin-top: 1rem; margin-bottom: 1rem; padding-bottom: 0.5rem;">
  If you use the `eyeris` package in your research, please cite it!
  
  Run the following in R to get the citation:
</div>

```{r}
citation("eyeris")
```