Group 8 Presentation

class: center, middle, inverse, title-slide

# Group 8 Presentation
### Alex Vand, Clarissa Boyajian, Scout Leonard
### 2021/10/27

---

#Our Question

**How do population demographic factors impact lead exposure in Philadelphia?**

- Developed from an interest in an EJ question
- DataONE had environmental health data for this question specifically from a 2021 published study

---

#Data Management Plan

- **Data managers:** Alex Vand, Clarissa Boyajian, and Scout Leonard
- **Repository owner:** Scout Leonard (Clarissa and Alex are collaborators)
- This team effort calls two datasets: one which contains **lead risk factors for Philadelphia communities**, and another which contains **human demographic data.** 
  - 6 hours for the 2 datasets
  - Code to call/clean/combine data in our repo
- **Preservation plan:** We plan to save until the end of MEDS!
- No legal constraints.

---

#Obtaining and Merging Data

- Data Retrieval
  - Public Health/Lead Data: `metajam` package
    - This uses an API to download the DataONE data
    - This creates the `metajam` log; stashed in our `.gitignore`
  - Census/demographic data: `censusapi` package
    - Requires an access token to control server traffic
    - `censusapi` package restricted us from pulling based on 1 county; we downloaded all of the data from state 42 before filtering
- Data Combination
  - Left joined datasets based on census tract
  - Census tract format required cleaning in lead dataset

---

#Analysis and Results

---

#Analysis and Results

- Winsorized to control for outliers

```r
median_income_95 <- quantile(
  lead_census_joined$acs_median_income_2019,
  probs = 0.95, na.rm = TRUE)

lead_census_joined <- lead_census_joined %>% 
  mutate(income_winsor = case_when(
    acs_median_income_2019 >= median_income_95 ~ 
      median_income_95,
    acs_median_income_2019 < median_income_95 ~ 
      acs_median_income_2019))

model <- lm(number_of_children_with_ebll_2015 ~ income_winsor,
            data = lead_census_joined)
model %>% summary()
```

---

#Analysis and Results

- OLS results showing 5% variation in EBLL is explained by median income

---

#Future Analysis

- Compare **multiple socioeconomic factors** to individual lead risk factors or resulting health impacts
- Spatial visualization of these variables
- The real world is more complicated!

---

##Data Preservation: KNB and Github

- GitHub
  - ReadMe with Data Management Plan
  - Code to call original datasets
  - Code for combining datasets 
  - Analyses
- KNB
  - Our final combined dataset in .csv format
  - Metadata for our combined dataset 
  - Our contact info and Orcid IDs
  
<img src="images/knb.png" width="3824" />