Analyzing an experiment on involuntary attention using brms

class: center, middle, inverse, title-slide

# Analyzing an experiment on <br /> involuntary attention using <strong><code>brms</code></strong>
### Antonio Schettino<br /> <br /> Department of Experimental-Clinical & Health Psychology<br /> Ghent University

---

layout: true

<div class="my-header"></div>

<div class="my-footer"><span>Antonio Schettino   
&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;
&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;
&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;
&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;&emsp;
Bayes@Lund - 12.04.2018</span></div>

---
name: outline

.left[
  .font200[
    **OUTLINE**
  ]
]

<br />

.font120[
  * background
]

.font120[
  * experimental paradigm
]

.font120[
  * data and results
]

.font120[
  * conclusions
]

--

<Div style="margin-top:-380px" />

.right[
  ![collaborators](collaborators.jpg)
]

---
name: notification_trolling

.center[
  .font200[**Involuntary attention**
  ]
]

<br />

.center[
  <body>
  <video width="320" height="240" controls="controls" vid.autoplay="false">
  <source src="notification_troll.mp4" type="video/mp4">
  <object data="" width="320" height="240">
  <embed width="320" height="240" src="notification_troll.mp4">
  </object></video></body>
  <br /><br />
  source: https://www.youtube.com/watch?v=0M2L9XNYfLs
]

---
name: clarify_example1

.center[
  .font200[**What happened?**
  ]
]

.font120[
  * The guy in the background was attracted to the phone...
]

.font120[
  * ... despite *knowing* that his friend was making the notification sounds
]

.font120[
  * His attention was **automatically** attracted to ~~uninformative~~ **counterproductive** sounds (distraction from current task: watching TV)
]

---
name: clarify_example2

.center[
  .font200[**Why study involuntary attention?**
  ]
]

.font120[
  * Involuntary attentional orienting can be dangerous in some real-life situations
]

.font120[
  *  Car driver distracted by flashing mobile phone, worker operating heavy machinery distracted by blinking lights, ...
]

.font120[
  * How to study this phenomenon in the lab?
]

---
name: TOJ

.center[
  .font200[**Temporal Order Judgment (TOJ)**
  ]
]

.font120[
  * Flash two stimuli on screen
]

.font120[
  * Task: which stimulus appeared **first**?
]

.font120[
  * Difficulty depends on the time between the onset of the stimuli (**SOA**)
]

---
name: TOJ_example
background-image: url("TOJ_video.gif")
background-size: 700px
background-position: 50% 65%

.center[
  .font200[**Temporal Order Judgment (TOJ)**
  ]
]

---
name: TOJ_cue

.center[
  .font200[**Counterproductive TOJ**
  ]
]

.font120[
  * An **exogenous cue** is used to attract attention towards one placeholder
]

.font120[
  * The stimulus on the attended location is perceived as **first** even when appearing second
]

.font120[
  * What if the cue is **always wrong**, i.e., appearing on the location of the *second* stimulus?
]

---
name: TOJ_video_cue
background-image: url("TOJ_video_cue.gif")
background-size: 700px
background-position: 50% 80%

.center[
  .font200[**Counterproductive TOJ**
  ]
]

---
name: TOJ_graph

.center[
  .font200[**TOJ - Data**
  ]
]

.center[
![](Schettino_BayesAtLund2018_files/figure-html/TOJ_plot-1.svg)
]

---
name: TOJ_analysis_desc1

.center[
  .font200[**Bayesian multilevel modeling**
  ]
]

---
name: TOJ_analysis_desc2

.center[
  .font200[**Bayesian multilevel modeling**
  ]
]

.font120[
* logistic regression, varying intercepts & slopes on *participants*
]

---
name: TOJ_analysis_desc3

.center[
  .font200[**Bayesian multilevel modeling**
  ]
]

.font120[
* logistic regression, varying intercepts & slopes on *participants*
* **highly informative priors** (from a pilot study)
]

---
name: TOJ_analysis_desc4

.center[
  .font200[**Bayesian multilevel modeling**
  ]
]

.font120[
* logistic regression, varying intercepts & slopes on *participants*
* **highly informative priors** (from a pilot study)
* model comparison:
  1. full model (**SOA** + **cue** + **SOA** x **cue**)
  2. main effects (**SOA** + **cue**)
  3. main effect of **SOA**
  4. main effect of **cue**
  5. **null** model (intercept only)
]

---
name: TOJ_analysis_desc5

.center[
  .font200[**Bayesian multilevel modeling**
  ]
]