diverging_bar_graphs.qmd

# Diverging bar graphs {#sec-diverging-bar-graphs}

```{r}
#| label: setup
#| message: false
#| warning: false
#| include: false
library(tidyverse)
library(lubridate)
library(scales)
library(knitr)
library(kableExtra)
library(colorblindr)
library(downlit)
# fonts ----
library(extrafont)
library(sysfonts)
source("_common.R")
# use font
showtext::showtext_auto()
# set theme
ggplot2::theme_set(theme_ggp2g(
    base_size = 14))
```

```{r}
#| label: status
#| results: "asis"
#| echo: false
# status ----
# polish, dev, draft, complete
status("complete")
```


:::: {.callout-note collapse="false" icon=false}

<br>

```{r}
#| label: full_code_display
#| eval: true
#| echo: false
#| warning: false
#| message: false
#| out-height: '60%'
#| out-width: '60%'
#| fig-align: right
library(fivethirtyeight) 
library(ggplot2)
library(patchwork)

fivethirtyeight::trump_approval_trend |> 
  filter(subgroup == "All polls") |> 
  mutate(
    month = lubridate::month(modeldate,
                label = TRUE, abbr = TRUE),
    approve = approve_estimate*0.01,
    disapprove = disapprove_estimate*0.01,
    disapprove = disapprove * -1) |> 
  pivot_longer(cols = c(approve, disapprove), 
    names_to = "poll", values_to = "values") |> 
  group_by(month, poll) |> 
    summarise(
      month_avg = mean(values, na.rm = TRUE)
    ) |> 
  ungroup() -> trump_approval_diverg

labs_geom_bar_diverg <- labs(
  title = "Trump Approval Ratings",
  subtitle = "From 'How Popular is Donald Trump'",
  x = "Month",
  y = "Monthly average percent",
  fill = "Estimate")
ggp2_bars_diverg <- ggplot(
  data = trump_approval_diverg, 
    aes(x = month, y = month_avg)) +
  geom_bar(aes(fill = poll),
    stat = "identity", width = .5) + 
  scale_y_continuous(limits = c(-1, 1), 
    labels = scales::percent) 

ggp2_bars_diverg + 
  labs_geom_bar_diverg +
  theme_ggp2g(
    base_size = 9)
```

::: {style="font-size: 1.10em; color: #02577A;"}
**This graph requires:**
:::

::: {style="font-size: 1.05em; color: #282b2d;"}
**`r emo::ji("check")` a categorical variable**      
:::

::: {style="font-size: 1.05em; color: #282b2d;"}
**`r emo::ji("check")` a numeric (continuous) variable**
:::

::::

## Description

If you have two proportions that contain positive and negative values, consider using diverging bars with `geom_bar()`. 

Unlike a standard or stacked bar graphs, diverging bar graphs display positive and negative quantities on both sides of a reference or baseline value (zero in this example). Color, length and position are used to compare the quantities across categorical levels (and within variable values). 

### Horizontal bars

```{r}
#| label: diverg
#| eval: true
#| echo: false
#| warning: false
#| message: false
#| out-height: '75%'
#| out-width: '75%'
#| fig-align: center
labs_geom_bar_diverg <- labs(
  x = "Month",
  y = "Monthly Avg percent",
  fill = "Estimate")
ggp2_bars_diverg <- ggplot(
  data = trump_approval_diverg, 
    aes(x = month, y = month_avg)) +
  geom_bar(aes(fill = poll),
    stat = "identity", width = .5) + 
  scale_y_continuous(limits = c(-1, 1), 
    labels = scales::percent) + 
    theme(legend.position = "bottom")
ggp2_bars_diverg + 
  labs_geom_bar_diverg + 
    theme_ggp2g(base_size = 11)
```

For example, we can use the length of the bar from the reference line to compare disapproval estimates across all months (i.e., comparing red bars to each other). 

### Vertical bars

```{r}
#| label: diverg_vert
#| eval: true
#| echo: false
#| warning: false
#| message: false
#| out-height: '75%'
#| out-width: '75%'
#| fig-align: center
labs_geom_bar_diverg_vert <- labs(
  x = "Monthly Avg percent",
  y = "Month",
  fill = "Estimate")

ggp2_bar_diverg_vert <- ggplot(
  data = trump_approval_diverg,
      aes(x = month_avg, y =  month)) +
  geom_bar(
      aes(fill = poll), 
        stat = "identity", width = .5) + 
  scale_x_continuous(limits = c(-1, 1), 
        labels = scales::percent)
ggp2_bar_diverg_vert + 
  labs_geom_bar_diverg_vert + 
    theme_ggp2g(base_size = 11)
```

We can also compare approval vs. disapproval for each month (i.e., compare the blue vs. red bars to each other within each month).

## Set up


::: {style="font-size: 1.15em; color: #1e83c8;"}
**PACKAGES:**
:::


Install packages.


::: {style="font-size: 0.85em;"}
```{r}
#| label: pkg_code_diverging_bar
#| code-fold: show
#| eval: true
#| echo: true
#| warning: false
#| message: false
#| results: hide
install.packages("fivethirtyeight")
library(fivethirtyeight) 
library(ggplot2)
```
:::



::: {style="font-size: 1.15em; color: #1e83c8;"}
**DATA:**
:::

::: {.column-margin}
![](www/538.png){fig-align="right" width="45%" height="45%"}
:::


Create `trump_approval_diverg` from the `trump_approval_trend` dataset in the `fivethirtyeight` package.

::: {style="font-size: 0.85em;"}
```{r}
#| label: data_code_diverging_bar
#| code-fold: show
#| eval: true
#| echo: true
fivethirtyeight::trump_approval_trend |> 
  dplyr::filter(subgroup == "All polls") |> 
  dplyr::mutate(
    month = lubridate::month(modeldate,
                label = TRUE, abbr = TRUE),
    approve = approve_estimate*0.01,
    disapprove = disapprove_estimate*0.01,
    disapprove = disapprove * -1) |> 
  tidyr::pivot_longer(cols = c(approve, disapprove), 
    names_to = "poll", values_to = "values") |> 
  dplyr::group_by(month, poll) |> 
    dplyr::summarise(
      month_avg = mean(values, na.rm = TRUE)
    ) |> 
  dplyr::ungroup() -> trump_approval_diverg
glimpse(trump_approval_diverg)
```
:::


## Grammar

::: {style="font-size: 1.15em; color: #1e83c8;"}
**CODE:**
:::

  - Create labels with `labs()`

  - Initialize the graph with `ggplot()` and provide `data`

  - Map the `month` to the `x` and `month_avg` to the `y`

  - Inside` geom_bar()`
  
    - map `poll` to `fill`
  
    - use `stat = "identity"` and `width = .5`
  
  - Add `scale_y_continuous()` to manually set the limits and format the axis with `scales::percent`

::: {style="font-size: 0.85em;"}
```{r}
#| label: code_graph_diverging_bar
#| code-fold: show
#| eval: false
#| echo: true 
#| warning: false
#| message: false
#| out-height: '100%'
#| out-width: '100%'
labs_geom_bar_diverg <- labs(
  title = "Trump Approval Ratings",
  subtitle = "From 'How Popular is Donald Trump'",
  x = "Month",
  y = "Monthly average percent",
  fill = "Estimate")
ggp2_bars_diverg <- ggplot(
  data = trump_approval_diverg, 
    aes(x = month, y = month_avg)) +
  geom_bar(aes(fill = poll),
    stat = "identity", width = .5) + 
  scale_y_continuous(limits = c(-1, 1), 
    labels = scales::percent) 

ggp2_bars_diverg + 
  labs_geom_bar_diverg
```
:::



::: {style="font-size: 1.15em; color: #1e83c8;"}
**GRAPH:**
:::

```{r}
#| label: create_graph_diverging_bar
#| eval: true
#| echo: false
#| warning: false
#| message: false
#| out-height: '100%'
#| out-width: '100%'
#| layout-nrow: 1
#| column: page-inset-right
labs_geom_bar_diverg <- labs(
  title = "Trump Approval Ratings",
  subtitle = "From 'How Popular is Donald Trump'",
  x = "Month",
  y = "Monthly average percent",
  fill = "Estimate")
ggp2_bars_diverg <- ggplot(
  data = trump_approval_diverg, 
    aes(x = month, y = month_avg)) +
  geom_bar(aes(fill = poll),
    stat = "identity", width = .5) + 
  scale_y_continuous(limits = c(-1, 1), 
    labels = scales::percent)

ggp2_bars_diverg + 
  labs_geom_bar_diverg
```

## More info

For vertically arranged bars, we switch the `x` and `y` axis variables.

### Vertically arranged bars

<br>

  - Create labels with `labs()`

  - Map the `month_avg` to the `x` and `month` to the `y` 

  - *Inside* `geom_bar()` 

    - map `poll` to `fill`  

    - use `stat = "identity"` and `width = .5`  
  
  - Add `scale_y_continuous()` to manually set the limits and format the axis with `scales::percent` 

::: {style="font-size: 0.85em;"}

```{r}
#| label: create_graph_vert_diverg_bars
#| eval: true
#| echo: true
#| warning: false
#| message: false
#| out-height: '100%'
#| out-width: '100%'
#| layout-nrow: 1
#| column: page-inset-right
labs_geom_bar_diverg_vert <- labs(
  title = "Trump Approval Ratings",
  subtitle = "From 'How Popular is Donald Trump'",
  x = "Monthly average percent",
  y = "Month",
  fill = "Estimate")
ggp2_bar_diverg_vert <- ggplot(
  data = trump_approval_diverg,
      aes(x = month_avg, y =  month)) +
  geom_bar(
      aes(fill = poll), 
        stat = "identity", width = .5) + 
  scale_x_continuous(limits = c(-1, 1), 
        labels = scales::percent)
ggp2_bar_diverg_vert + 
  labs_geom_bar_diverg_vert
```

:::