Data Visualization

class: center, middle, inverse, title-slide

.title[
# Data Visualization
]
.subtitle[
## What does the data look like?
]
.author[
### Prof. Weldzius
]
.institute[
### Villanova University
]

---

# Agenda

1. Quick refresher on last lectures

2. Plotting in `R`

- `ggplot` (`+` instead of `%>%`)

---

# Loading Packages & Data

- Create an `.Rmd` file and save to your `code` folder

- Accept defaults, Save As... (with a good name), then `knit`

- Load the `tidyverse` package

``` r
require(tidyverse)
```

- Load the data from the course [github page](https://github.com/rweldzius/PSC4175/raw/main/static/data/sc_debt.Rds) directly using `read_rds()`

- We **create** an "object" to store the data using a left-arrow: `<-`

``` r
df<-read_rds("https://github.com/rweldzius/PSC4175/raw/main/static/data/sc_debt.Rds") 
```

---

# The Two Camps

---

# The .blue[Research] Camp

- .blue[RQ]: How might admissions and SAT scores be **related**?

- .blue[Theory]: selective schools have stricter criteria
  
--

- .blue[Hypothesis]: admissions and SAT scores should be **negatively** related
  
--

- How can we test this hypothesis?

---

# Previously: `summarise()`

- We can combine base `R` functions with `tidyverse` functions!

- Base `R`: `mean()`
  
  - `tidyverse`: `summarise()` (aka `summarize()`)
  
--
  
- Overall average SAT scores

``` r
df %>%
  summarise(mean_sat = mean(sat_avg,na.rm=T))
```

```
## # A tibble: 1 × 1
##   mean_sat
##      <dbl>
## 1    1141.
```

---

# Previously: `summarise()`

- Let's unpack this

``` r
df %>%
  summarise(mean_sat = mean(sat_avg,na.rm=T))
```

- Create new variable `mean_sat` that contains the `mean()` of every school's average SAT score
  
--

- `na.rm=T` means we want to ignore missing data. If not?
  
--

``` r
df %>%
  summarise(mean_sat = mean(sat_avg))
```

```
## # A tibble: 1 × 1
##   mean_sat
##      <dbl>
## 1       NA
```

---

#`summarise()` + `filter()`

- Recall we want see if more selective schools have higher SAT scores

``` r
df %>%
  filter(adm_rate < .1) %>%
  summarise(mean_sat_LT10 = mean(sat_avg,na.rm=T))
```

```
## # A tibble: 1 × 1
##   mean_sat_LT10
##           <dbl>
## 1         1510.
```

``` r
df %>%
  filter(adm_rate > .1) %>%
  summarise(mean_sat_GT10 = mean(sat_avg,na.rm=T))
```

```
## # A tibble: 1 × 1
##   mean_sat_GT10
##           <dbl>
## 1         1135.
```

---

# `summarise()` + `group_by()`

- One final `tidyverse` function: `group_by()`

``` r
df %>%
  group_by(selective) %>%
  summarise(mean_sat = mean(sat_avg,na.rm=T))
```

```
## # A tibble: 3 × 2
##   selective mean_sat
##       <dbl>    <dbl>
## 1         0    1135.
## 2         1    1510.
## 3        NA     NaN
```

---

# Plotting data

- Let's plot the data instead of writing many of these `summarise()` functions

- Visualization in `R` uses `ggplot()` function

- Inputs: `aes(x,y,...)` (elipses `...` indicates many more inputs)
  
  - `x` is the x-axis (horizontal)
  
  - `y` is the y-axis (vertical)

---

# `ggplot()`

- Attach `ggplot()` to your data with `%>%`

``` r
*df %>%
  ggplot()
```

---

# `ggplot()`

- Then tell it what to put in the x-axis and y-axis

- What should go on these axes?

- .blue[Theory]: Selective schools choose higher scoring students

- Selective schools **explain** higher scores
  
  - Selective schools: **independent variable** / **explanatory variable** / **predictor** / `\(X\)`
  
  - Higher scores: **dependent variable** / **outcome variable** / `\(Y\)`
  
--

- Selective schools go on the x-axis, SAT scores go on the y-axis

---

# `ggplot()`

``` r
df %>%
* ggplot(aes(x = adm_rate,y = sat_avg))
```

---

# `ggplot()`

- This gives us an empty plot

- We have the correct variables on the correct axes...

- ...but we need to choose how to display them

- There are many different `ggplot()` functions to choose from

- `geom_point()` creates one point for each x and y coordinate
  
  - `geom_bar()` creates a barplot
  
  - `geom_histogram()` creates a histogram
  
  - `geom_density()` creates a density plot
  
  - `geom_boxplot()` creates a box-and-whisker plot
  
---

# `ggplot()`

- We **add** a second `ggplot()` function to the first with a plus sign `+`

- **NB:** This is JUST LIKE THE PIPE OPERATOR `%>%` in `tidyverse`!
  
--

- Since `adm_rate` (the x-axis variable) and `sat_avg` (the y-axis variable) are both numeric ("continuous") measures, we will use `geom_point()`

- We will come back to **variable types** and how to visualize them later
  
---

# `ggplot()`

``` r
df %>%
  ggplot(aes(x = adm_rate,y = sat_avg)) + 
* geom_point()
```

---

# Plotting data

- Let's unpack this

- `aes(x,y)` sets the basic aesthetics for the plot
  
--

- `geom_point()` tells `ggplot()` how to visualize those aesthetics
  
--

- These two parts are linked with the `+`. Similar to...?

- ...the `%>%` in `tidyverse`!

---

# Interpreting the plot

- We **hypothesized** that admissions and SAT scores are negatively related

- Is this supported in the data?
  
--

- Let's add a line of best fit with `geom_smooth()`

``` r
df %>%
  ggplot(aes(x = adm_rate,y = sat_avg)) + 
  geom_point() + 
* geom_smooth(method = 'lm',se = F)
```

---

---

# The .blue[Research] Camp

- .blue[RQ]: How might future earnings and SAT scores be **related**?

- .blue[Theory]: SATs measure student ability.
  
  - .blue[Theory]: Student ability is valued by the labor market.
  
  - .blue[Theory]: Firms pay more for students with higher SAT scores.
  
--

- .blue[Hypothesis]: Earnings and SAT scores should be **positively** related
  
---

# Plotting Quiz

- Which variable goes on the x-axis?

- **SAT scores**
  
--

- Which variable goes on the y-axis?

- **Earnings**
  
--

- In our theory, SAT scores **cause** earnings

- Why might this **not** be the case?

- Spurious 1: SAT scores **and** earnings are caused by student ability
  
  - Spurious 2: SAT scores **and** earnings are caused by socio-economic privilege
  
---

# Let's Plot!

``` r
df %>%
  ggplot(aes(x = sat_avg,y = md_earn_wne_p6)) + # Build axes
  geom_point() +  # Add points
  geom_smooth(method = 'lm',se = F) # Add line of best fit
```

<img src="3_DataViz_files/figure-html/unnamed-chunk-16-1.png" style="display: block; margin: auto;" />
---

# Outliers

- Which schools are furthest from the line?

- These are **outliers**
  
--

- These schools are the **furthest** from our .blue[theory]

``` r
df %>%
  mutate(out = ifelse(md_earn_wne_p6 > 100000,
                          instnm,  # Value if TRUE
                          NA)) %>% # Value if FALSE
  drop_na(out,sat_avg) %>%
  select(instnm,md_earn_wne_p6,sat_avg)
```

```
## # A tibble: 2 × 3
##   instnm                              md_earn_wne_p6 sat_avg
##   <chr>                                        <int>   <int>
## 1 University of Health Sciences and …         120400    1262
## 2 Albany College of Pharmacy and Hea…         112100    1242
```

---

# Plotting data

- We can add these as labels!

``` r
df %>%
   mutate(out = ifelse(md_earn_wne_p6 > 100000,
                          instnm,  # Value if TRUE
                          NA)) %>% # Value if FALSE
  ggplot(aes(x = sat_avg,y = md_earn_wne_p6,
*            label = out)) +
  geom_point() + 
* geom_text()
```

---

# Plotting data

---

# Plotting data

- Let's accentuate the outlier more with color

``` r
p <- df %>%
   mutate(out = ifelse(md_earn_wne_p6 > 100000,
                          instnm,  # Value if TRUE
                          NA)) %>% # Value if FALSE
  drop_na(sat_avg) %>%
  ggplot(aes(x = sat_avg,y = md_earn_wne_p6,
*            label = out,color = out)) +
  geom_point() + 
* scale_color_manual(name = "Outlier",values = c('red','red','black')) +
  geom_text(hjust = .5,vjust = 1,color = 'black',size = 3)
```

---

# Plotting data

---

# Categorical Data

- Thus far, plotting two continuous variables with `geom_point()`

- What if we wanted to see which state has the most selective schools?

- Use `group_by()` and `summarise()`
  
--

``` r
df %>%
  group_by(stabbr) %>%
  summarise(selective_avg = mean(adm_rate,na.rm=T))
```

```
## # A tibble: 51 × 2
##    stabbr selective_avg
##    <chr>          <dbl>
##  1 AK             0.827
##  2 AL             0.654
##  3 AR             0.676
##  4 AZ             0.843
##  5 CA             0.592
##  6 CO             0.768
##  7 CT             0.589
##  8 DC             0.529
##  9 DE             0.627
## 10 FL             0.620
## # ℹ 41 more rows
```

---

# Categorical Data

- Two variables (`stabbr` and `selective_avg`), but one of them is now a `character` type

- Can we plot this as a scatterplot?

``` r
p <- df %>%
  group_by(stabbr) %>%
  summarise(selective_avg = mean(adm_rate,na.rm=T)) %>%
  ggplot(aes(x = stabbr,y = selective_avg)) + 
  geom_point()
```

---

# Categorical Data

- Yes...but it isn't very pretty

``` r
p
```

---

# Categorical Data: `geom_bar()`

- NB: `geom_bar()` will automatically count the values on the x-axis
  
--

``` r
df %>%
  ggplot(aes(x = stabbr)) + 
  geom_bar()
```

---

# Categorical Data: `geom_bar()`

- This is fine if we just want to know which states have the most schools in our data

- But we want to put the average admissions rate on the y-axis instead

- Need to **override** `geom_bar()` default behavior

``` r
p <- df %>%
  group_by(stabbr) %>%
  summarise(selective_avg = mean(adm_rate,na.rm=T)) %>%
  ggplot(aes(x = stabbr,y = selective_avg)) + 
* geom_bar(stat = 'identity')
```

---

# Categorical Data

``` r
p
```

---

# Categorical Data

- Getting a little better, but still ugly

- Use `reorder()` to sort the x-axis values by the y-axis

``` r
p <- df %>%
  group_by(stabbr) %>%
  summarise(selective_avg = mean(adm_rate,na.rm=T)) %>%
* ggplot(aes(x = reorder(stabbr,selective_avg),y = selective_avg)) +
  geom_bar(stat = 'identity')
```

---

# Categorical Data

- Even better!

``` r
p
```

---

# Plot Tweaking

- We could go even further and swap the x and y-axes (although this isn't always a good idea!)

``` r
p <- df %>%
  group_by(stabbr) %>%
  summarise(selective_avg = mean(adm_rate,na.rm=T)) %>%
* ggplot(aes(y = reorder(stabbr,selective_avg),x = selective_avg)) +
  geom_bar(stat = 'identity')
```

---

# Plot Tweaking

``` r
p
```

- Still ugly though! We want to tweak the labels with `labs()`

---

# Plot Tweaking

``` r
p + 
  labs(title = "Average Admissions by State",
       x = "Admissions Rate",
       y = "State")
```

---

# Conclusion

- What to take away

1. Which variables go on which axes
  
  2. How to put these on a `ggplot()` figure
  
  3. How to create a visualization of these variables