Vectors

c(1, 3, 5)

c(TRUE, FALSE, TRUE, TRUE)

c("red", "blue")

Vectors

Vectors

Vectors have 1 dimension

Vectors

Vectors have 1 dimension

Vectors have a length.

length(c("blue", "red"))

Vectors

Vectors have 1 dimension

Vectors have a length.

length(c("blue", "red"))

Some vectors have names.

names(c("x" = 1, "y = 1))

Vectors

Vectors have 1 dimension

Vectors have a length.

length(c("blue", "red"))

Some vectors have names.

names(c("x" = 1, "y = 1))

Vectors have types

Types

Numeric/double

Integer

Factor

Character

Logical

Dates

Packages to work with types:

Strings/character: stringr

Packages to work with types:

Strings/character: stringr

Factors: forcats

Packages to work with types:

Strings/character: stringr

Factors: forcats

Dates: lubridate

Making vectors

1:3

## [1] 1 2 3

c(1, 2, 3)

## [1] 1 2 3

rep(1, 3)

## [1] 1 1 1

seq(from = 1, to = 3, by = .5)

## [1] 1.0 1.5 2.0 2.5 3.0

Your Turn 1

Create a character vector of colors using c(). Use the colors "grey90" and "steelblue". Assign the vector to a name.

Use the vector you just created to change the colors in the plot below using scale_color_manual(). Pass it using the values argument.

Your Turn 1

cols <- c("grey90", "steelblue")
gapminder %>% 
  mutate(rwanda = ifelse(country == "Rwanda", TRUE, FALSE)) %>% 
  ggplot(aes(year, lifeExp, color = rwanda, group = country)) + 
  geom_line() +
  scale_color_manual(values = cols) +
  theme_minimal()

Your Turn 1

Working with vectors

Subset vectors with [] or [[]]

x <- c(1, 5, 7)

x[2]

## [1] 5

x[[2]]

## [1] 5

x[c(FALSE, TRUE, FALSE)]

## [1] 5

Working with vectors

Modify elements

x

## [1] 1 5 7

x[2] <- 100

x

## [1]   1 100   7

Modify elements

x

## [1]   1 100   7

x[x > 10] <- NA

x

## [1]  1 NA  7

cols <- c("grey90", "steelblue")
gapminder %>% 
  mutate(rwanda = ifelse(country == "Rwanda", TRUE, FALSE)) %>% 
  ggplot(aes(year, lifeExp, color = rwanda, group = country)) + 
  geom_line() +
  scale_color_manual(values = cols) +
  theme_minimal()

cols <- c("grey90", "steelblue") 
gapminder %>% 
  mutate(rwanda = ifelse(country == "Rwanda", TRUE, FALSE)) %>% 
  ggplot(aes(year, lifeExp, group = country)) +
  geom_line( 
    data = function(x) filter(x, !rwanda),
    color = cols[1]
  ) + 
  theme_minimal()

cols <- c("grey90", "steelblue") 
gapminder %>% 
  mutate(rwanda = ifelse(country == "Rwanda", TRUE, FALSE)) %>% 
  ggplot(aes(year, lifeExp, color = rwanda, group = country)) + 
  geom_line(
    data = function(x) filter(x, !rwanda), 
    color = cols[1]
  ) +
  geom_line(
    data = function(x) filter(x, rwanda),
    color = cols[2],
    size = 1.5
  ) + 
  theme_minimal()

Your Turn 2

Create a numeric vector that has the following values: 3, 5, NA, 2, and NA.

Try using sum(). Then add na.rm = TRUE.

Check which values are missing with is.na(); save the results to a new object and take a look

Change all missing values of x to 0

Try sum() again without na.rm = TRUE.

Your Turn 2

x <- c(3, 5, NA, 2, NA)
sum(x)

## [1] NA

Your Turn 2

sum(x, na.rm = TRUE)

## [1] 10

Your Turn 2

x_missing <- is.na(x)
x_missing

## [1] FALSE FALSE  TRUE FALSE  TRUE

x[x_missing] <- 0
x

## [1] 3 5 0 2 0

sum(x)

## [1] 10

Writing functions

Writing functions

Writing functions

Writing functions

Your Turn 3

Create a function called sim_data that doesn't take any arguments.

In the function body, we'll return a tibble.

For x, have rnorm() return 50 random numbers.

For sex, use rep() to create 50 values of "male" and "female". Hint: You'll have to give rep() a character vector. for the first argument. The times argument is how many times rep() should repeat the first argument, so make sure you 3. account for that.

For age() use the sample() function to sample 50 numbers from 25 to 50 with replacement.

Call sim_data()

Your Turn 3

sim_data <- function() {
  tibble(
    x = rnorm(50), 
    sex = rep(c("male", "female"), times = 25),
    age = sample(25:50, size = 50, replace = TRUE)
  )
}
sim_data()

Your Turn 3

sim_data <- function() {
  tibble(
    x = rnorm(50),
    sex = rep(c("male", "female"), times = 25),
    age = sample(25:50, size = 50, replace = TRUE)
  )
}
sim_data()

Your Turn 3

sim_data <- function() {
  tibble(
    x = rnorm(50), 
    sex = rep(c("male", "female"), times = 25),
    age = sample(25:50, size = 50, replace = TRUE)
  )
}
sim_data()

Your Turn 3

sim_data <- function() {
  tibble(
    x = rnorm(50), 
    sex = rep(c("male", "female"), times = 25),
    age = sample(25:50, size = 50, replace = TRUE)
  )
}
sim_data()

Your Turn 3

## # A tibble: 50 x 3
##           x sex      age
##       <dbl> <chr>  <int>
##  1  0.312   male      42
##  2 -0.387   female    25
##  3 -0.0210  male      38
##  4  1.38    female    33
##  5  0.796   male      30
##  6 -0.996   female    29
##  7 -0.442   male      38
##  8  0.00711 female    28
##  9  1.16    male      45
## 10  0.116   female    49
## # … with 40 more rows

E-Values

The strength of unmeasured confounding required to explain away a value

E-Values

The strength of unmeasured confounding required to explain away a value

Rate ratio: 3.9 = E-value: 7.3

Your Turn 4

Write a function to calculate an E-Value given an RR.

Call the function evalue and give it an argument called estimate. In the body of the function, calculate the E-Value using estimate + sqrt(estimate * (estimate - 1))

Call evalue() for a risk ratio of 3.9

Your Turn 4

evalue <- function(estimate) {
  estimate + sqrt(estimate * (estimate - 1))
}

evalue(3.9)

## [1] 7.263034

Control Flow

if (PREDICATE) {
  true_result
}

if (PREDICATE) {
  true_result
} else {
  default_result
}

if (PREDICATE) {
  true_result
} else if (ANOTHER_PREDICATE) {
  true_result
} else  {
  default_result
}

Other functions to control flow

ifelse(PREDICATE, true_result, false_result)
dplyr::case_when(
  PREDICATE ~ true_result,
  PREDICATE ~ true_result, 
  TRUE ~ default_result
)
switch(
  x,
  value1 = result,
  value2 = result
)

Validation and stopping

if (is.numeric(x))

stop(), warn()

function(x) {
  if (is.numeric(x)) stop("x must be a character")
  # do something with a character
}

Your Turn 5

Use if () together with is.numeric() to make sure estimate is a number. Remember to use ! for not.

If the estimate is less than 1, set estimate to be equal to 1 / estimate.

Call evalue() for a risk ratio of 3.9. Then try 0.80. Then try a character value.

Your Turn 5

evalue <- function(estimate) { 
  if (!is.numeric(estimate)) stop("`estimate` must be numeric")
  if (estimate < 1) estimate <- 1 / estimate
  estimate + sqrt(estimate * (estimate - 1))
}

Your Turn 5

evalue(3.9)

## [1] 7.263034

evalue(.80)

## [1] 1.809017

evalue("3.9")

## Error in evalue("3.9"): `estimate` must be numeric

Your Turn 6

Add a new argument called type. Set the default value to "rr"

Check if type is equal to "or". If it is, set the value of estimate to be sqrt(estimate)

Call evalue() for a risk ratio of 3.9. Then try it again with type = "or".

Your Turn 6

evalue <- function(estimate, type = "rr") {
  if (!is.numeric(estimate)) stop("`estimate` must be numeric")
  if (type == "or") estimate <- sqrt(estimate)
  if (estimate < 1) estimate <- 1 / estimate
  estimate + sqrt(estimate * (estimate - 1))
}

Your Turn 6

evalue(3.9)

## [1] 7.263034

evalue(3.9, type = "or")

## [1] 3.362342

Your Turn 7: Challenge!

Create a new function called transform_to_rr with arguments estimate and type.

Use the same code above to check if type == "or" and transform if so. Add another line that checks if type == "hr". If it does, transform the estimate using this formula: (1 - 0.5^sqrt(estimate)) / (1 - 0.5^sqrt(1 / estimate)).

Move the code that checks if estimate < 1 to transform_to_rr (below the OR and HR transformations)

Return estimate

In evalue(), change the default argument of type to be a character vector containing "rr", "or", and "hr".

Get and validate the value of type using match.arg(). Follow the pattern argument_name <- match.arg(argument_name)

Transform estimate using transform_to_rr(). Don't forget to pass it both estimate and type!

Your Turn 7: Challenge!

transform_to_rr <- function(estimate, type) {
  if (type == "or") estimate <- sqrt(estimate)
  if (type == "hr") { 
    estimate <- 
      (1 - 0.5^sqrt(estimate)) / (1 - 0.5^sqrt(1 / estimate)) 
  } 
  if (estimate < 1) estimate <- 1 / estimate
  estimate
}
evalue <- function(estimate, type = c("rr", "or", "hr")) {
  # validate arguments
  if (!is.numeric(estimate)) stop("`estimate` must be numeric")
  type <- match.arg(type) 
  # calculate evalue
  estimate <- transform_to_rr(estimate, type)
  estimate + sqrt(estimate * (estimate - 1))
}

Your Turn 7: Challenge!

transform_to_rr <- function(estimate, type) { 
  if (type == "or") estimate <- sqrt(estimate)
  if (type == "hr") {
    estimate <-
      (1 - 0.5^sqrt(estimate)) / (1 - 0.5^sqrt(1 / estimate))
  }
  if (estimate < 1) estimate <- 1 / estimate
  estimate
}
evalue <- function(estimate, type = c("rr", "or", "hr")) {
  # validate arguments
  if (!is.numeric(estimate)) stop("`estimate` must be numeric")
  type <- match.arg(type) 
  # calculate evalue
  estimate <- transform_to_rr(estimate, type) 
  estimate + sqrt(estimate * (estimate - 1))
}

Your Turn 7: Challenge!

transform_to_rr <- function(estimate, type) { 
  if (type == "or") estimate <- sqrt(estimate)
  if (type == "hr") { 
    estimate <- 
      (1 - 0.5^sqrt(estimate)) / (1 - 0.5^sqrt(1 / estimate)) 
  } 
  if (estimate < 1) estimate <- 1 / estimate
  estimate
}
evalue <- function(estimate, type = c("rr", "or", "hr")) {
  # validate arguments
  if (!is.numeric(estimate)) stop("`estimate` must be numeric")
  type <- match.arg(type)
  # calculate evalue
  estimate <- transform_to_rr(estimate, type) 
  estimate + sqrt(estimate * (estimate - 1))
}

Your Turn 7: Challenge!

evalue(3.9)

## [1] 7.263034

evalue(3.9, type = "or")

## [1] 3.362342

evalue(3.9, type = "hr")

## [1] 4.474815

evalue(3.9, type = "rd")

## Error in match.arg(type): 'arg' should be one of "rr", "or", "hr"

Pass the dots: ...

select_gapminder <- function(...) {
  gapminder %>% 
    select(...)
}
select_gapminder(pop, year)

Pass the dots: ...

select_gapminder <- function(...) {
  gapminder %>% 
    select(...)
}
select_gapminder(pop, year)

Pass the dots: ...

## # A tibble: 1,704 x 2
##         pop  year
##       <int> <int>
##  1  8425333  1952
##  2  9240934  1957
##  3 10267083  1962
##  4 11537966  1967
##  5 13079460  1972
##  6 14880372  1977
##  7 12881816  1982
##  8 13867957  1987
##  9 16317921  1992
## 10 22227415  1997
## # … with 1,694 more rows

Your Turn 8

Use ... to pass the arguments of your function, filter_summarize(), to filter().

In summarize, get the n and mean life expectancy for the data set

Check filter_summarize() with year == 1952.

Try filter_summarize() again for 2002, but also filter countries that have "and" in the country name. Use str_detect() from the stringr package.

Your Turn 8

filter_summarize <- function(...) {
  gapminder %>% 
    filter(...) %>%
    summarize(n = n(), mean_lifeExp = mean(lifeExp))
}

filter_summarize(year == 1952)

## # A tibble: 1 x 2
##       n mean_lifeExp
##   <int>        <dbl>
## 1   142         49.1

filter_summarize(year == 2002, str_detect(country, " and "))

## # A tibble: 1 x 2
##       n mean_lifeExp
##   <int>        <dbl>
## 1     4         69.9

Programming with dplyr, ggplot2, and friends

plot_hist <- function(x) {
  ggplot(gapminder, aes(x = x)) + geom_histogram()
}

Programming with dplyr, ggplot2, and friends

plot_hist <- function(x) {
  ggplot(gapminder, aes(x = x)) + geom_histogram()
}

plot_hist(lifeExp)

## Error in FUN(X[[i]], ...): object 'lifeExp' not found

Programming with dplyr, ggplot2, and friends

plot_hist <- function(x) {
  ggplot(gapminder, aes(x = x)) + geom_histogram()
}

plot_hist("lifeExp")

## Error: StatBin requires a continuous x variable: the x variable is discrete.Perhaps you want stat="count"?

Curly-curly

plot_hist <- function(x) {
  ggplot(gapminder, aes(x = {{x}})) + geom_histogram()
}

Curly-curly

plot_hist <- function(x) {
  ggplot(gapminder, aes(x = {{x}})) + geom_histogram()
}

plot_hist(lifeExp)

Your turn 9

Filter gapminder by year using the value of .year (notice the period before hand!). You do NOT need curly-curly for this. (Why is that?)

Arrange it by the variable. Don't forget to wrap it in curly-curly!

Make a scatter plot. Use variable for x. For y, we'll use country, but to keep it in the order we arranged it by, we'll turn it into a factor. Wrap the the factor() call with fct_inorder(). Check the help page if you want to know more about what this is doing.

Your turn 9

top_barplot <- function(variable, .year) {
  gapminder %>%
    filter(year == .year) %>%
    arrange(desc({{variable}})) %>% 
    #  take the 10 lowest values
    tail(10) %>% 
    ggplot(aes(x = {{variable}}, y = fct_inorder(factor(country)))) +
      geom_point() +
      theme_minimal()
}

Your turn 9

top_barplot <- function(variable, .year) {
  gapminder %>%
    filter(year == .year) %>% 
    arrange(desc({{variable}})) %>%
    #  take the 10 lowest values
    tail(10) %>% 
    ggplot(aes(x = {{variable}}, y = fct_inorder(factor(country)))) +
      geom_point() +
      theme_minimal()
}

Your turn 9

top_barplot(lifeExp, 2002)

Your turn 9

top_barplot(lifeExp, 2002) + 
  labs(x = "Life Expectancy", y = "Country")