gistfile1.txt

---
title: "Preparing Experimental Data for Analysis"
subtitle: "R + Quarto RevealJS"
author: "Elena Peden, Lylla Turco, Chloe Lockwood, Julianna Ross"
date: "3/23/26"
format:
  revealjs:
    theme: simple
    template-partials: 
      - title-slide.html
    css: pooh-theme.css
    slide-number: true
    incremental: true
    center: false
    transition: fade
    code-line-numbers: true
execute:
  echo: true
  message: false
  warning: false
---

## Attentional Blink

- Attentional blink is a brief lapse in attention that occurs after a person detects one target stimulus, making it harder to notice a second target that follows shortly afterward. 
- In rapid streams of visual information, attention becomes temporarily occupied with processing the first target, so the second target is often missed if it appears too soon.
- As the time between the two targets increases, it is generally expected that detection rates of the second target will improve

------------------------------------------------------------------------

## Instructions

:::: incremental
::: {style="font-size: 0.55em;"}
-   A sequence of 19 letters appeared, with each new letter replacing the previous one
-   Each letter was shown for 100 milliseconds
-   Participants waited until the sequence ended before responding
-   Participants reported whether they saw:
    -   ::::: columns
        ::: {.column width="50%"}
        -   J

        -   K
        :::

        ::: {.column width="50%"}
        -   Both

        -   Neither
        :::
        :::::
-   The experiment included 60 total trials Procedures:
-   Each trial presented a sequence of 19 letters at 100 ms per letter
-   The sequence may or may not have included the target letters J and/or K
-   Participants responded after each sequence indicating which targets they saw
-   The position of J and K varied across trials
-   The key focus was how accurately participants detected the second target letter
-   Specifically, accuracy was compared based on the number of letters between the first and second targets
:::
::::

------------------------------------------------------------------------

## Data Pipeline

- This data from this task gives us evidence for the "Attentional Blink" which is the phenomenon in which there is a brief time after paying attention to one stimulus when attention cannot be focused on a subsequent stimulus.

- While we know the data shows this, we can't tell that just from looking at thousands of data points. (111,246 to be exact) That's where the data pipeline comes in.

## Data Pipeline

This process follows a structure of Rearranging ⇝ Cleaning/Conditioning ⇝ Exploring ⇝ Comprehending

::::: columns
::: {.column width="50%" style="font-size: 0.75em"}
-   Rearranging: convert raw logs to tidy trial rows
-   Cleaning: remove or repair invalid observations
-   Conditioning: apply exclusion rules tied to design
-   Exploring: check distributions, outliers, and group patterns
-   Comprehending: interpret patterns and assess model readiness
:::

::: {.column width="50%"}
Like Pooh bear, we are trying to find the honey in our raw data honeycomb.

![](Img/honey_in_the_honeycomb.jpeg)
:::
:::::

------------------------------------------------------------------------

## Raw Data Structure

-   Columns in raw export:
    -   `Trial`
    -   `First Target`
    -   `Second Target`
    -   `Separation`(Independent Variable)
    -   `Response`(Dependent Variable)
    -   `randgator` (Participant ID)

------------------------------------------------------------------------

## Prepping the Data

Goal: set ourselves up for success with coding later

```{r, echo = FALSE}
library("tidyverse") 
library("forcats")
library("janitor")

data_file <-"attentional_blink_rgtr.csv"

data_set <- read_csv(data_file)
```

```{r}

attentional_blink <- data_set %>% clean_names() 

```

Example: First Target ➙ first_target

- This formats data so there are no capitals, spaces, etc.

------------------------------------------------------------------------

## Recoding

**First:** simplify the labels

```{r}
lutboth<-c("K only"="K","Both"="KJ","J only"="J","Neither"="")
```

ex: K only ⇾ K

**Second:** apply to the table

```{r}
attentional_blink$targ_resp<-lutboth[attentional_blink$response]
```

------------------------------------------------------------------------

## Score the Data

**Goal:** Make a new data frame with binary values so data can be plotted

```{r}
ab_per<-attentional_blink %>% rowwise %>%  
  mutate(first = 
           if_else(str_sub(targ_resp,1)==first_target,1,0),
         second = 
           if_else(str_sub(targ_resp,-1)==second_target,1,0))
```

This code does 2 things:

-   `first` : compares the first character of the response to the first target

-   `second` : compares the last character of the response to the second target

------------------------------------------------------------------------

## Trouble with Both

-   In the earlier code, we can have a problem when the participant recognizes both stimuli
-   Both is coded as KJ but what happens if J was the first target?
-   J gets compared to K and is scored `0` and K is compared to J and scored `0` as well
-   We need to override that code

------------------------------------------------------------------------

## Trouble with Both

------------------------------------------------------------------------

## Exploration

```{r}

```

------------------------------------------------------------------------

## Interpretation

-   Where are the largest deviations from expected behavior?
-   Do we see an attentional blink pattern by separation?
-   Are there participants with unusual response profiles?

------------------------------------------------------------------------

## Model Compatibility

-   Data now satisfy key modeling assumptions:
    -   One row per trial
    -   Consistent factor levels
    -   Exclusions documented and reproducible
    -   Derived variables aligned to design

------------------------------------------------------------------------

## Takeaways

-   Most research effort is in preprocessing
-   Conditioning decisions shape inference
-   Quarto + R makes this workflow explicit and reproducible