Visualizing Missing Data

class: center, middle, inverse, title-slide

.title[
# Visualizing Missing Data
]
.subtitle[
## <br><br> College of the Atlantic
]
.author[
### Laurie Baker
]

---

class: middle

# Visualizing Missing Data

---
## Origins of Missing Data

- What are the origins of missing data?

---
## Packages for visualizing missing data

```r
library(visdat)
library(naniar)
library(UpSetR)
```

---
## Inspecting the data types

Let's explore a data set of imaginary people who have been randomly generated.

```r
glimpse(typical_data)
```

```
## Rows: 5,000
## Columns: 9
## $ ID           <chr> "0001", "0002", "0003", "0004", "0005", "0…
## $ Race         <fct> Black, Black, Black, Hispanic, NA, White, …
## $ Age          <dbl> NA, 25, 31, 27, 21, 22, 23, 21, NA, 27, 22…
## $ Sex          <fct> Male, Male, Female, Female, Female, Female…
## $ `Height(cm)` <dbl> 175.9, 171.7, 173.5, 172.4, 158.5, 169.5, …
## $ IQ           <dbl> 110, 84, 115, 84, 116, 83, 101, 97, 92, 99…
## $ Smokes       <lgl> FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, F…
## $ Income       <dbl> 91, 877, 3190, 4239, 3995, 557, 273, 1133,…
## $ Died         <lgl> FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, F…
```
* Which of the variable's data types have likely been incorrectly identified?

.footnote[For more info about the randomly generated data run `?typical_data`]

---
## Summarizing Missing data

```r
summary(typical_data)
```

```
##       ID                   Race           Age       
##  Length:5000        White    :2938   Min.   :20.00  
##  Class :character   Hispanic : 745   1st Qu.:23.00  
##  Mode  :character   Black    : 525   Median :27.00  
##                     Asian    : 181   Mean   :27.47  
##                     Bi-Racial:  74   3rd Qu.:31.25  
##                     (Other)  :  37   Max.   :35.00  
##                     NA's     : 500   NA's   :500    
##      Sex         Height(cm)          IQ          Smokes       
##  Male  :2499   Min.   :140.3   Min.   : 60.0   Mode :logical  
##  Female:2501   1st Qu.:168.8   1st Qu.: 93.0   FALSE:4085     
##                Median :175.2   Median :100.0   TRUE :915      
##                Mean   :175.2   Mean   :100.1                  
##                3rd Qu.:181.8   3rd Qu.:107.0                  
##                Max.   :208.9   Max.   :137.0                  
##                                NA's   :500                    
##      Income        Died        
##  Min.   :   1   Mode :logical  
##  1st Qu.:1126   FALSE:2474     
##  Median :2250   TRUE :2526     
##  Mean   :2250                  
##  3rd Qu.:3374                  
##  Max.   :4499                  
##  NA's   :500
```
* `summary` can give us some ideas of missing data

---
## Visualizing Data Types

```r
vis_dat(typical_data)
```

---
## Visualizing Missing Data

```r
vis_miss(typical_data)
```

---
## Calculating how much is missing

```r
# From nanianr package:

n_miss(typical_data) # number missing
```

```
## [1] 2000
```

```r
n_complete(typical_data) # number complete
```

```
## [1] 43000
```

```r
pct_miss(typical_data) # percent missing
```

```
## [1] 4.444444
```

---
# Calculating how much is missing by column

```r
miss_var_summary(typical_data)
```

```
## # A tibble: 9 × 3
##   variable n_miss pct_miss
##   <chr>     <int>    <dbl>
## 1 Race        500       10
## 2 Age         500       10
## 3 IQ          500       10
## 4 Income      500       10
## 5 ID            0        0
## 6 Sex           0        0
## # ℹ 3 more rows
```

---
# Visualizing missing data by variable

```r
gg_miss_var(typical_data)
```

---
# Visualizing percent missing data by variable

```r
gg_miss_var(typical_data, show_pct = TRUE) +
  ylim(0, 100)
```

---
# Visualizing missing data by another variable

```r
gg_miss_var(typical_data,
            facet = Race)
```

---
# Missingness across factors

```r
gg_miss_fct(x = typical_data, fct = Race) +
  labs(y = "Variable")
```

---
# Missingness across cases

```r
gg_miss_case(typical_data)
```

---
# Missingness across cases

```r
typical_data %>%
  as_shadow_upset() %>%
  upset()
```

---
# Daily Air Quality Data in NYC, May-Sept 1973

```r
vis_dat(airquality)
```

---
# Returning to the `ggplot` warnings

```r
ggplot(airquality, 
       aes(x = Solar.R, y = Ozone)) + 
  geom_point()
```

```
## Warning: Removed 42 rows containing missing values
## (`geom_point()`).
```

---
# Relationships in missing values

```r
airquality %>%
  ggplot(aes(x = Ozone, y = Solar.R)) +
  geom_miss_point()
```

---
# Relationships in missing values

```r
airquality %>%
  ggplot(aes(x = Ozone, y = Solar.R)) +
  geom_miss_point() +
  facet_wrap(~Month)
```

---
# Visualizing relationships in missing variables

```r
airquality %>%
    bind_shadow() %>%
    ggplot(aes(x = Solar.R, fill = Ozone_NA)) +
  geom_histogram(binwidth = 20)
```

```
## Warning: Removed 7 rows containing non-finite values
## (`stat_bin()`).
```

---
# Consequences of Missing Data

- So you have data that is missing? What now?

---
# Your Turn

Your Turn:

<a href="https://commons.wikimedia.org/wiki/File:Sasquatch.svg">Happybluemo</a>, <a href="https://creativecommons.org/licenses/by-sa/4.0">CC BY-SA 4.0</a>, via Wikimedia Commons

---
# Thank you