Einführung in R (Frühjahrssemester 2024): Tag 3 - Recap

# Skript zu Tag 3 - Wiederholungsfragen Data Prep

The notes made on the questions that you posed during day 3 on the content from day 2.

Day3-Questions-on-Day2.R

here::i_am("R/Day3-Questions-on-Day2.R")
library(here)
library(magrittr)
library(dplyr)
library(tidyr)
library(data.table)

# These are the notes taken during the third day, when we discussed questions
#  on the topics from the second day.

# Using pipes--------------------------

# The data from which to start:
gdp <- DataScienceExercises::aggGDPlifexp
head(gdp)

## Comparing code with and without pipes---------

# Restructure the data WITHOUT pipes:
gdp_2 <- tidyr::pivot_longer(
  data = gdp, 
  cols = -c("continent",  "year"), 
  names_to = "variable", 
  values_to = "observation")
head(gdp_2)

gdp_3 <- tidyr::pivot_wider(
  data = gdp_2, 
  names_from = "continent", 
  values_from = "observation")
head(gdp_3)

# Restructure the data WITH pipes:
gdp_final <- gdp %>% 
  tidyr::pivot_longer(
    cols = -c("continent",  "year"), 
    names_to = "variable", 
    values_to = "observation"
    ) %>% 
  tidyr::pivot_wider(
    names_from = "continent", 
    values_from = "observation")

## Using the new pipe: |> --------------
#  '|>' is largely equivalent to '%>%' and available in base R, but only in more
#  recent version. See for more details:
#  https://www.tidyverse.org/blog/2023/04/base-vs-magrittr-pipe/
gdp_final <- gdp |>
  tidyr::pivot_longer(
    cols = -c("continent",  "year"), 
    names_to = "variable", 
    values_to = "observation"
  ) |>
  tidyr::pivot_wider(
    names_from = "continent", 
    values_from = "observation")


## When using the . with the pipe makes sense------
# The pipe throws the result of one line into the next line; if you do not
#  want to use the result as input to the first argument, you can use .
#  to use it for a different argument. Here is one (somehow artificial) example
#  on when this could be useful, using the data from the exercises:
join_x <- fread(here("data/raw/join_x.csv"))
join_y <- fread(here("data/raw/join_y.csv"))

joined_data <- join_y %>% 
  filter(id=="b") %>% 
  left_join(
    x = join_x, 
    y = ., # <- here comes the intermediate output passed by the pipe
    by=c("time", "id"))

# Alternative without the pipe:
join_y_2 <- filter(join_y, id=="b")
joined_data <- left_join(
    x = join_x, 
    y = join_y_2, 
    by=c("time", "id"))



# Selection helpers--------------------
beer_data <- DataScienceExercises::beer
head(beer_data)

# Select the columns manually:
beer_data_price <- beer_data %>% 
  select(c("price", "price_liquor", "price_other"))
head(beer_data_price)

# Select the columns using selection helpers:
beer_data_price <- beer_data %>% 
  select(starts_with("pri"))
head(beer_data_price)

# You can also combine several selection helper within c()
beer_data_price_2 <- beer_data %>% 
  select(c(starts_with("pri"), ends_with("ion")))
head(beer_data_price_2)
beer_data_price_2

# And selection helpers can also be used within many other functions from 
#  the tidyverse:

beer_data %>% 
  pivot_longer(
    cols = contains("price"), 
    names_to = "price_based")

# Using selection helpers within the across() function
# You can compute the means for the three price-based columns manually:
beer_data %>% 
  summarize(
    p=mean(price),
    p_l=mean(price_liquor),
    p_o=mean(price_other)
  )

# Or you can use selection helpers within across():
beer_data %>% 
  summarise(across(.cols = contains("price"), .fns = mean))
# You can find more information on across here:
# https://dplyr.tidyverse.org/reference/across.html

# General questions--------------------
## Different column names when joining-----------
# Can I join data sets on columns that have different names in the two data
#  sets that should be joined?

# Download example data:
wdi_gdp <- WDI::WDI(
  country = c("DE", "AT", "IT"), 
  start = 2000, end = 2020, 
  indicator = c("GDP"="NY.GDP.PCAP.PP.KD", 
                "Oil_Rent"="NY.GDP.PETR.RT.ZS"))

# Create example data:
wdi_gdp_1 <- wdi_gdp %>% 
  select(country, year, GDP)
wdi_gdp_2 <- wdi_gdp %>% 
  select(country, year, Oil_Rent) %>% 
  rename(Country=country)

wdi_rejoined <- full_join(
  x = wdi_gdp_1, y = wdi_gdp_2, 
  by = c("country"="Country", "year")
  )
# Hint: look at the package countrycode, which is very useful when working
#  with country codes:
#  https://github.com/vincentarelbundock/countrycode

## Selecting many years-----------
# How do I best use many years when filtering?

# This works but is quite complex:
wdi_gdp_filtered <- wdi_gdp %>% 
  filter(year %in% c(2010, 2011, 2012, 2014, 2015, 2016), 
         country=="Italy" | country=="Germany")

# This is a more straightforward variant that does the same:
wdi_gdp_filtered <- wdi_gdp %>% 
  filter(year %in% c(2010:2012, 2014:2016),
         country %in% c("Italy", "Germany"))

# Some examples on how the %in% command works:
let <- letters[1:4]
let

"A" %in% let # Is "A" in the vector let?
"d" %in% let # Is "d" in the vector let?

# Related are the set operators:

set_1 <- 1:4
set_2 <- 3:6

union(set_1, set_2) # All elements that are at least in one of the two sets
intersect(set_1, set_2) # Elements that are in both sets
setdiff(set_1, set_2) # Elements that are in set_1, but not set_2
setdiff(set_2, set_1) # Order matters: elements that are in set_2, but not set_1
setequal(set_1, set_2) # Do the two sets contain the same elements?
setequal(set_1, 1:4) # It is about the same elements, not whether it is the same set!