Weekly Update for June 25

# Weekly Update for June 25
### 2018/06/25

---

---
class: primary 
# New CSAFE slide template

We're now using [`xaringan`](https://github.com/yihui/xaringan)

What's changed:

- New person slide:

````
---
class: inverse
# Your Name
````

---
class: secondary

- New content slide with title:

````
---
class: primary
# Title of slide

Slide content
````

- New content slide without title:

````
---
class: secondary

Slide content with no title on slide
````

---
class: inverse
# Sample User

---
class: primary
# Sample Slide

- Sample Table:

| Team     | Group | W   | D   | L   |
| : ------ | :---- | :-- | :-- | :-- |
| Germany  | F     | 1   | 0   | 1   |
| Mexico   | F     | 2   | 0   | 0   |
| Sweden   | F     | 1   | 0   | 1   |
| S. Korea | F     | 0   | 0   | 2   |

---
class: primary
# Sample Slide

- Sample Graphic:

---
class: secondary

- If you are going to add an image, **create a directory** with your name within "images" folder. For example, "images/**guillermo**/sample_image.JPG"

<img src = "images/guillermo/sample_image.JPG" width="35%"/>
---

---

# SLRs

- Reading about ABC to be able to understand the papers Chris recommended to me. 
- Assuming no mistakes, `$e^{E[log(SLR)|H_0]} \leq E\left[\frac{SLR}{LR} | H_0 \right]$`.
    - left side is computable
    - right side is some measure of the average quality of the LR under `$H_0$` 
    - still not that useful for bounding probabilities

---
class: secondary

<<<<<<< HEAD
$$
`\begin{aligned}
=======
`$$\begin{aligned}
>>>>>>> 95be075b1c4a221ae2d3e915c591938b79d6d1c1
E\left[\frac{SLR}{LR}|H_0\right] &= e^{log\left( E\left[\frac{SLR}{LR} | H_0 \right] \right)} \\
& \geq e^{E\left[ log\left(\frac{SLR}{LR} \right) |H_0 \right]} \\
& = e^{E\left[ log(SLR) |H_0 \right] - E\left[ log(LR) |H_0 \right]} \\
& \geq e^{E\left[ log(SLR) |H_0 \right] - log\left(E\left[ LR |H_0 \right] \right)} \\
& = e^{E\left[ log(SLR) |H_0 \right]}
<<<<<<< HEAD
\end{aligned}`
$$
=======
\end{aligned}$$`
>>>>>>> 95be075b1c4a221ae2d3e915c591938b79d6d1c1
---
class: primary

# Groove ID

- Reading about changepoint detection
- Possible approach similar to some examples given in "Bayesian Retrospective Multiple-Changepoint Identification" (Stephens 1994) in JRSSc. 
    - Doesn't seem fundamentally tied to time series data

---
class: inverse  
# Kiegan

---
class: primary  
# What I'm Working On

- I have been banned from working on my paper and the book until after the prelim  
- Prelim is in two weeks!

---
class: inverse
<<<<<<< HEAD
=======
# Amy

---
class: primary
# What I've been up to:
      
- May 31 (talk): The ABA's Ninth Annual Prescription for Criminal Justice Forensics Program in New York

- June 5 (outreach): 4th and 5th graders for a STEM camp (thanks Jimmy!)

- June 7 and 8 (outreach): STEMversity with middle school students in Milledgeville, GA

- June 11-12 (meetings and poster): All Hands. Handwriting pre-meeting with the Israei team and Dr. Stern went well.
      
---
class: primary
# Next:  
      
<br />  
<br />

- Prelim study (next 2 weeks)

- Handwriting after that (finally)

- July 31 (talk - handwriting): JSM
- August 19-23 (poster? talk? maybe?): The American Society of Questioned Document Examiners (ASQDE)
 
 
---
class: inverse
>>>>>>> 95be075b1c4a221ae2d3e915c591938b79d6d1c1
# Jimmy

---
class: primary
# Longitudinal Data Collection
## Final Collection 
* Issues with the 2D Scanner 
    + Phantom Images
    + Theoretically, wont take long
* Possible Causes
    + Past cleaning Procedures
    + Shoes (from being cleaned)
    + Finger Print Powder
* "Z" Drive 
    + Reviewing Errors (5438)

---
class: primary
# This Week
## **Moving Forward**
* Adapt to the Problem
     + Continue Cleaning and figure it out
     + Finish Data Collection
<<<<<<< HEAD
=======

---
class:inverse
# Ben and Jenny

---
class:primary
# Working on

## Shoe Tread Classification

* 668 shoes partially/fully classified as of 10am

## Longitudinal shoe project

* Helped Miranda scan shoes Thursday/Friday

---
class:inverse
# Susan

---
class:primary
# Shoe Tread

* (2 weeks ago) set up LabelMe docker container to classify shoes

![LabelMe Interface](images/susan/LabelMe.png)

---
class:primary
# Shoe Tread
* (Last week) Code to "carve up" tread annotations into useful pictures (using Matlab toolbox from LabelMe)

* Pictures tend to look better if the regions are square-ish
    
<img src="images/susan/line-1.jpg" width = 30%/>

---
class:primary
# Truthiness

* "When making rapid judgments about the truth of a claim, nonprobative images lead people to believe the claim"
    - Want to know if this holds for charts and graphs
    - Some evidence that brain images cause the person presenting the images to appear more credible in court

* Finding claims that are data-driven and can be made into a chart is hard.

---
class:primary
# Truthiness
* First step - pilot study with ~ 10 claims    
Variables: 
    - Related/Not Related to claim
    - Probative/Non-probative
    - Level of Abstraction
        - High: statistical/data chart
        - Low: map/"infographic" that uses "common knowledge" as a baseline
        - Image (no data at all)
        
Prediction: more abstraction = lower familiarity, so non-probative abstract images will produce less of an effect.

---
class:inverse
# Miranda

---
class:primary
# Updates

- 2D shoe scanning
- Starting neural network training soon with labeled images
  - Thanks, Jenny and Ben!
- Brainstorming facts for truthiness

---
class:inverse
# Heike

---
class:primary
# The Hamby Set that wasn't

* Background: NIST has two sets of Hamby bullets 44 and 252
* ... or that's what we thought ...

---
class: inverse
# Sam

---
class: primary
# Happy Birthday!

Happy Birthday Malisha!!!

![](images/sam/malishaAllHands.png)
>>>>>>> 95be075b1c4a221ae2d3e915c591938b79d6d1c1