2023-w02-birds.qmd

Set working directory

```{r}
setwd(here::here('2023-w02-birds'))
```




Load packages and data

```{r}
library(tidyverse)

feederwatch <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2023/2023-01-10/PFW_2021_public.csv')
site_data <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2023/2023-01-10/PFW_count_site_data_public_2021.csv')
```


Clean names.
This makes it easier to work with column names.

```{r}
feederwatch_clean <- feederwatch |> janitor::clean_names()
site_data_clean <- site_data |> janitor::clean_names()
```

Take a look at column names and compare with data dictionary

```{r}
colnames(feederwatch_clean)
colnames(site_data_clean)
```


It looks like there are many 'fed_in' variable names in the 'site_data' data set.
Let's take a look at all of them.
Tidyselect helpers will give us a selection.

```{r}
site_data_clean |> 
  select(starts_with('fed'))
```


This looks weird.
It's only zeroes and ones and NAs.
Probably a true/false kind of thing.

Let's bring more columns into this.
There's `loc_id` and `proj_period_id`.

```{r}
site_data_clean |> 
  select(loc_id, proj_period_id, starts_with('fed'))
```


This is starting to make sense.
Each feeding site has a unique location and a project id that contains what looks like a year.
Let's check how many project IDs there are.

```{r}
unique(site_data_clean$proj_period_id)
```

All project IDs contain the same prefix.
Let's remove it and transform the character vector into an actual numeric vector.
`parse_number()` can take care of that.

```{r}
parse_number(site_data_clean$proj_period_id)[1:10]
```

Perfect, now let's save this data set.

```{r}
sites_fed <- site_data_clean |> 
  select(loc_id, proj_period_id, starts_with('fed')) |> 
  mutate(year = parse_number(proj_period_id), .before = 2) |> 
  select(-proj_period_id)
sites_fed
```


Next, we're going to take care of missing values.
Let's have a look how many missing values there are.

::: panel-tabset

## Using `across()`

```{r}
sites_fed |> 
  summarise(across(.cols = everything(), .fns = ~sum(is.na(.)))) 
```

## Using for-loop

```{r}
columns <- colnames(sites_fed)
missing_vals <- seq_along(columns)
names(missing_vals) <- columns

for (col in columns) {
  missing_vals[col] <- sum(is.na(sites_fed[[col]])) 
}
missing_vals
```

:::

There is missing data.
Let's filter those that have missing data in any of the month columns.
The fed_yr_round column can be filled by us then.

::: panel-tabset

## Functional programming `{purrr}`

```{r}
complete_monthly_infos <- sites_fed |> 
  drop_na(fed_in_jan:fed_in_dec) |> 
  mutate(across(-c(loc_id, year), as.logical)) 

complete_monthly_infos$fed_yr_round <- pmap_lgl(
  .l = complete_monthly_infos |> select(-c(loc_id, year, fed_yr_round)), 
  .f = all
)
```

## Using `rowSums()`

```{r}
complete_monthly_infos <- sites_fed |> 
  drop_na(fed_in_jan:fed_in_dec) |> 
  mutate(across(-c(loc_id, year), as.logical)) 

number_of_months_fed <- complete_monthly_infos |> 
    select(-c(loc_id, year, fed_yr_round)) |> 
    rowSums()

complete_monthly_infos$fed_yr_round <- (number_of_months_fed == 12)

```

:::


Now, let us bring our data into a tidy format.
That's what `pivot_longer()` will do for us.

```{r}
sites_fed_infos <- complete_monthly_infos |> 
  pivot_longer(
    cols = -c(loc_id, year),
    names_to = 'month',
    names_prefix = 'fed_in_',
    values_to = 'fed'
  )
sites_fed_infos
```

Next, we're able to do a little bit of counting.

```{r}
fed_counts <- sites_fed_infos |> 
  count(year, month, fed)
fed_counts
```

Let's check how many sites there are over the years.

```{r}
sites_over_years <- fed_counts |> 
  filter(month != 'fed_yr_round') |> 
  group_by(year) |> 
  summarise(n = sum(n))

sites_over_years |> 
  ggplot(aes(year, n)) + 
  geom_line()
```

Looks like overall the number of sites increased over the years.
This plot was just something we did for ourselves.
No need to customize it further.

Finally, let's have a look at how many sites feed all-year.
Maybe over time more or maybe less bird sites are active every month.

```{r}
sites_fed_infos |> 
  filter(month == 'fed_yr_round') |> 
  ggplot(aes(x = year, fill = fed)) +
  geom_bar(position = 'fill')
```

Alright, it looks like there is a trend that more and more bird sites are active every month.
Let's make this viz a bit prettier.

First, let's apply `theme_minimal()` and make the bars wider (plus black border).

```{r}
sites_fed_infos |> 
  filter(month == 'fed_yr_round') |> 
  ggplot(aes(x = year, fill = fed)) +
  geom_bar(position = 'fill', col = 'black', width = 1) +
  theme_minimal(base_size = 14) 
```

Second, add labels. 
Don't forget to put your Twitter handle into the caption.

```{r}
sites_fed_infos |> 
  filter(month == 'fed_yr_round') |> 
  ggplot(aes(x = year, fill = fed)) +
  geom_bar(position = 'fill', col = 'black', width = 1) +
  theme_minimal(base_size = 14) +
  labs(
    x = element_blank(),
    y = 'Share of bird sites',
    fill = 'Feeds all-year',
    title = 'Over the years, the share of bird sites that feed every\nmonth of the year increased',
    caption = 'TidyTuesday 2023 - Week 02 | Viz: @rappa753'
  ) 
```

Third, let us format the y-axis as percent.

```{r}
sites_fed_infos |> 
  filter(month == 'fed_yr_round') |> 
  ggplot(aes(x = year, fill = fed)) +
  geom_bar(position = 'fill', col = 'black', width = 1) +
  theme_minimal(base_size = 14) +
  labs(
    x = element_blank(),
    y = 'Share of bird sites',
    fill = 'Feeds all-year',
    title = 'Over the years, the share of bird sites that feed every\nmonth of the year increased',
    caption = 'TidyTuesday 2023 - Week 02 | Viz: @rappa753'
  ) +
  scale_y_continuous(labels = scales::percent_format()) 
```


Fourth, let's pick better colors manually.

```{r}
sites_fed_infos |> 
  filter(month == 'fed_yr_round') |> 
  ggplot(aes(x = year, fill = fed)) +
  geom_bar(position = 'fill', col = 'black', width = 1) +
  theme_minimal(base_size = 14) +
  labs(
    x = element_blank(),
    y = 'Share of bird sites',
    fill = 'Feeds all-year',
    title = 'Over the years, the share of bird sites that feed every\nmonth of the year increased',
    caption = 'TidyTuesday 2023 - Week 02 | Viz: @rappa753'
  ) +
  scale_y_continuous(labels = scales::percent_format()) +
  scale_fill_manual(values = c('grey90', 'dodgerblue2')) 
```

Fifth, get rid of the extra spacing surrounding the bars.

```{r}
sites_fed_infos |> 
  filter(month == 'fed_yr_round') |> 
  ggplot(aes(x = year, fill = fed)) +
  geom_bar(position = 'fill', col = 'black', width = 1) +
  theme_minimal(base_size = 14) +
  labs(
    x = element_blank(),
    y = 'Share of bird sites',
    fill = 'Feeds all-year',
    title = 'Over the years, the share of bird sites that feed every\nmonth of the year increased',
    caption = 'TidyTuesday 2023 - Week 02 | Viz: @rappa753'
  ) +
  scale_y_continuous(labels = scales::percent_format()) +
  scale_fill_manual(values = c('grey90', 'dodgerblue2')) +
  coord_cartesian(expand = FALSE)
```

Finally, move the legend and title.

```{r}
sites_fed_infos |> 
  filter(month == 'fed_yr_round') |> 
  ggplot(aes(x = year, fill = fed)) +
  geom_bar(position = 'fill', col = 'black', width = 1) +
  theme_minimal(base_size = 14) +
  labs(
    x = element_blank(),
    y = 'Share of bird sites',
    fill = 'Feeds all-year',
    title = 'Over the years, the share of bird sites that feed every\nmonth of the year increased',
    caption = 'TidyTuesday 2023 - Week 02 | Viz: @rappa753'
  ) +
  scale_y_continuous(labels = scales::percent_format()) +
  scale_fill_manual(values = c('grey90', 'dodgerblue2')) +
  coord_cartesian(expand = FALSE) +
  theme(
    legend.position = 'top',
    plot.title.position = 'plot'
  )
```


There's lots more one can do with the data or the plot.
But as a start, this is probably okay.

For now, you can share your plot on Twitter using the #tidyTuesday hashtag.
If you do, think about sharing your code as well.
Common practices for sharing the code: A dedicated tidyTuesday repo on Github.
Or you can just upload the code at [gist.github.com](https://gist.github.com/).

If you want to learn more, then check out [https://www.rscreencasts.com/](https://www.rscreencasts.com/).
It's a project that documents all of [David Robinson](https://www.youtube.com/@safe4democracy/videos)'s great screencasts in which he's analyzing TidyTuesday data in lightning speed.