Practice for the practice Quiz

Using Problem 12.2.1 Exercise 2 as a guide, use the ideas from Chapter 13 to answer the questions for table2.

1. Compute the rate and include it in a final dataframe with the years as columns.

Answer:

The first answer approaches the problem by splitting the dataset into two and then joining the two dataset.

library(tidyverse)
table2

table2 %>% arrange(type)

table2_cases <- table2 %>% filter(type == "cases") %>% 
  select(country, year, count) %>%
  rename(cases = count)
table2_cases

library(stringr)
table2_pop <- table2 %>% filter(type == "population") %>% 
  select(country, year, count) %>%
  rename(population = count)
table2_pop

Now join the two datasets using two variables as the unique key.

table2_join <- table2_cases %>% inner_join(table2_pop, by=c("country", "year")) 
table2_join

Create the new column.

table2_new <- table2_join %>% mutate(rate = cases / population * 10000)
table2_new

Now spread the data out into two columns.

table2_new_spread <- table2_new %>% select(country, year, rate) %>%
  spread(year, rate)
table2_new_spread

Now try the new function pivot_wider(). Note new this function is from the tidyr 1.0 package.

table2_new_spread2 <- table2_new %>% select(country, year, rate) %>%
  pivot_wider(country, names_from = year, values_from = rate)
table2_new_spread2

Are the two files the same. Lets give the comparedf() function a try. It is from the arsenal R package.

library(arsenal)
comparedf(table2_new_spread, table2_new_spread2)

Compare Object

Function Call: 
comparedf(x = table2_new_spread, y = table2_new_spread2)

Shared: 3 non-by variables and 3 observations.
Not shared: 0 variables and 0 observations.

Differences found in 0/3 variables compared.
0 variables compared have non-identical attributes.

Anternative Solution:

Can we use spread from the beginning? Yes.

table2 %>% spread(key = type, value = count) %>%
  mutate(rate = cases/population) %>%
  select(-cases, -population) %>%
  spread(key = year, value = rate)

Or

table2 %>% pivot_wider(names_from = type, values_from = count) %>%
  mutate(rate = cases/population) %>%
  select(-cases, -population) %>%
  pivot_wider(names_from = year, values_from = rate)

2. Now make a clustered bar graph. Question, which table is the one to use, table2_new or table2_new_spread?

Answer: The one to use is in tidy format. So table2_new. Note the use of as.factor() function. This is our next topic of discussion.

table2_new %>% ggplot(aes(x = country, y = rate, fill = as.factor(year))) +
  geom_bar(stat = "identity", position = "dodge") +
  theme_light()

Or you can make the plot using year to group the bars.

table2_new %>% ggplot(aes(x = as.factor(year), y = rate, fill = country)) +
  geom_bar(stat = "identity", position = "dodge") +
  theme_light()

```