Becoming more efficient with functions and loop

• Automate common tasks
• Apply the same "procedure" to tens or hundreds of files, of objects, of dataframes, of variables.

• In a nutshell:
  • Functions allow to avoid copy pasting, allow to reuse a procedure
  • Loops allow to do one thing many times

• They are often used in combination.

Functions (justyfying their use)

set.seed(232)
df <- tibble::tibble(
  a = rnorm(10),
  b = rnorm(10),
  c = rnorm(10)
)
df$a <- (df$a - min(df$a, na.rm = TRUE)) / 
  (max(df$a, na.rm = TRUE) - min(df$a, na.rm = TRUE))
df$b <- (df$b - min(df$b, na.rm = TRUE)) / 
  (max(df$b, na.rm = TRUE) - min(df$b, na.rm = TRUE))
df$c <- (df$c - min(df$c, na.rm = TRUE)) / 
  (max(df$c, na.rm = TRUE) - min(df$c, na.rm = TRUE))
head(df)

## # A tibble: 6 × 3
##       a     b     c
##   <dbl> <dbl> <dbl>
## 1 1     0.277 0.667
## 2 0     0.375 0    
## 3 0.553 0.634 0.478
## 4 0.499 1     0.562
## 5 0.423 0.914 0.629
## 6 0.589 0.802 0.557

set.seed(232)
df <- tibble::tibble(
  a = rnorm(10),
  b = rnorm(10),
  c = rnorm(10)
)
rescale01 <- function(x) {
  rng <- range(x, na.rm = TRUE)
  (x - rng[1]) / (rng[2] - rng[1])
}
df$a <- rescale01(df$a)
df$b <- rescale01(df$b)
df$c <- rescale01(df$c)
head(df)

## # A tibble: 6 × 3
##       a     b     c
##   <dbl> <dbl> <dbl>
## 1 1     0.277 0.667
## 2 0     0.375 0    
## 3 0.553 0.634 0.478
## 4 0.499 1     0.562
## 5 0.423 0.914 0.629
## 6 0.589 0.802 0.557

Implementing the mean function by taking the sum, and dividing by the number of non missing cases.

my_vector <- c(3,8,19,NA,22,1)
mean(my_vector,na.rm=TRUE)

## [1] 10.6

custom_mean1 <- function(x){
  my_sum <- sum(x,na.rm=TRUE)
  my_length <- sum(!is.na(x))
  my_sum / my_length
}
custom_mean1(my_vector)

## [1] 10.6

Implementing the same mean function by going through all values one by one and calculating a bit like we would do in elementary school. This is often the logic in real functions. Breaking down everything.

custom_mean2 <- function(x){
  my_length <- 0
  my_sum <- 0
  for (i in my_vector){
    if (!is.na(i)){
      my_length <- my_length + 1
      my_sum <- my_sum + i
    }
  }
  my_sum / my_length
}
custom_mean2(my_vector)

## [1] 10.6

Using existing functions

Of course, you likely won't have to write many functions. A lot of functions exists. We've been using functions since the beginning of the semester.

mean() group_by() ggplot() (and many more) these are all functions.

Functions are incredibly useful.

You might, for example, want to:

• Calculate cross tabs and plot them, for many survey questions (à la what polling firms)
• Read in a court decision and clean it to a format you can analyze

Example

Imagine we have 200 respondents on 10 questions. Answers: {1:"Strongly disagree",2:"Somewhat disagree",3:"Neutral",4:"Somewhat agree",5:"Strongly agree",6:"Don't know"}. We want to plot the responses to each question in a separate figure.

questions <- paste0("question",1:10)
df <- as_tibble(matrix(sample(1:6,size=2000,replace = TRUE),nrow=200))
names(df) <- questions
df

## # A tibble: 200 × 10
##    question1 question2 question3 question4 question5 question6 question7
##        <int>     <int>     <int>     <int>     <int>     <int>     <int>
##  1         5         6         1         3         3         3         6
##  2         6         3         1         1         3         4         6
##  3         3         5         5         5         4         6         5
##  4         3         1         2         6         3         2         5
##  5         1         2         6         4         5         5         6
##  6         3         3         5         5         5         3         1
##  7         6         5         6         3         6         1         1
##  8         5         3         3         3         5         2         4
##  9         1         5         5         2         4         3         1
## 10         3         6         1         4         2         3         1
## # … with 190 more rows, and 3 more variables: question8 <int>, question9 <int>,
## #   question10 <int>

make_plot <- function(data,my_variable,answer_choices){
  to_plot <- data %>%
    select(all_of(my_variable)) %>%
    rename(selected_variable=all_of(my_variable)) %>%
    group_by(selected_variable) %>%
    count() %>%
    ungroup() %>%
    mutate(p=n/sum(n))  
  to_plot <- to_plot %>%
    mutate(selected_variable=recode(selected_variable,!!!answer_choices)) %>%
    mutate(selected_variable = factor(selected_variable,answer_choices))
  ggplot(to_plot,aes(x=selected_variable,y=p)) +
    geom_bar(stat='identity') +
    geom_text(aes(label = p), vjust = 1.5, colour = "white")
}

answer_choices=c("Strongly disagree","Somewhat disagree","Neutral",'Somewhat agree',"Strongly agree","Don't know")
make_plot(data=df,my_variable='question1',answer_choices=answer_choices)

make_plot(data=df,my_variable='question2',answer_choices)

make_data_to_plot <- function(data,my_variable){
  to_plot <- data %>%
    select(all_of(my_variable)) %>%
    rename(selected_variable=all_of(my_variable)) %>%
    group_by(selected_variable) %>%
    count() %>%
    ungroup() %>%
    mutate(p=n/sum(n))  
  to_plot <- to_plot %>%
    mutate(selected_variable=recode(selected_variable,!!!answer_choices)) %>%
    mutate(selected_variable = factor(selected_variable,answer_choices))
  to_plot
}
make_plot <-  function(data,answer_choices){
  ggplot(data,aes(x=selected_variable,y=p)) +
    geom_bar(stat='identity') +
    geom_text(aes(label = p), vjust = 1.5, colour = "white")
}

data_to_plot <- make_data_to_plot(data=df,my_variable='question2')
make_plot(data_to_plot,answer_choices)

Loops (Iteration chapter in the book)

threeletters <- c("a","b","c")
for (i in threeletters){
  print(i)
}

## [1] "a"
## [1] "b"
## [1] "c"

for (i in seq_along(threeletters)){
  print(i)
  print(threeletters[i])
}

## [1] 1
## [1] "a"
## [1] 2
## [1] "b"
## [1] 3
## [1] "c"

df <- tibble(
  a = rnorm(10),
  b = rnorm(10),
  c = rnorm(10),
  d = rnorm(10)
)

output <- vector("double", ncol(df))
for (i in seq_along(df)) {            
  output[[i]] <- median(df[[i]])      
}
output

## [1]  0.333332638  0.758161632 -0.002268765 -0.107977257

for (i in seq_along(df)) {            
  df[[i]] <- df[[i]]+10      
}
df

## # A tibble: 10 × 4
##        a     b     c     d
##    <dbl> <dbl> <dbl> <dbl>
##  1 10.9  10.6   9.47 10.8 
##  2  9.40  8.26  9.08 11.1 
##  3 10.3  11.0  11.0   8.72
##  4 11.5  11.7  12.0  10.2 
##  5 11.1  11.6  10.1  10.7 
##  6 10.2  11.1  10.3   9.57
##  7  9.23 10.4   9.93  9.02
##  8  9.88 11.3   9.56 11.2 
##  9 10.3  10.3  12.0   8.24
## 10 11.9   9.85  8.42  7.94