Session 3

Packages

How do you install a package from the R prompt, like readxl?

install.packages("readxl")

How do you load a package from the R prompt, like dplyr

library(dplyr)

How can you use RStudio to load a package?

Answer

RStudio Packages Pane

Okay, but what is a package?

Data Types

What data type are each of the following?

Run the following command. It will create 3 variables: x, y, and z. Without printing the variables, how can you tell what data type they are?

source("https://git.io/fNrm9")

x <- 1:10
y <- setNames(sample(letters, 10), LETTERS[1:10])
z <- runif(10)
z[sample(1:10, 2)] <- NA

y
z

class(x)
class(y)
class(z)

is.integer(x)
is.character(y)
is.na(z)

Workspaces & RStudio Projects

In the last session, we created an RStudio project and an example R script.

1. Re-open the project we created (or create a new project).
2. What is the current working directory?
3. Use the File pane to navigate to your desktop (or another folder on your computer).
4. How can you quickly navigate back to the working directory?

Open example_single_patient.R from our previous session and add the following lines.

example <- data_frame(
  patient_id = patient_id,
  age_dx = age_at_diagnosis,
  age_visit = age_at_visit,
  tumor_size = tumor_size,
  site_code = site_code
)

Clear your workspace (quick refresher here) and then source the script.

View the tibble that is stored in example.

# A tibble: 5 x 5
  patient_id age_dx age_visit tumor_size site_code
       <dbl>  <dbl>     <int>      <dbl> <chr>    
1    5554321     54        54        9.5 C220     
2    5554321     54        55        9.5 C400     
3    5554321     54        56        9.7 C412     
4    5554321     54        57        9.9 C220     
5    5554321     54        58       10.1 C400     

Functions that work with vectors

So far we’ve primarily seen vectors that operate on single values or that take single-valued arguments.

But as we’ve seen, R is a vectorized language.

Try using the following functions on the variables we created in example_single_patient.R in Session 2.

age_at_visit <- 54:58
tumor_size <- c(9.5, 9.5, 9.7, 9.9, 10.1)
site_code <- c("C220", "C400", "C412", "C220", "C400")

min(tumor_size)
max(tumor_size)
mean(tumor_size)
median(tumor_size)
var(tumor_size)
sd(tumor_size)
IQR(tumor_size)

Those functions all come from base R (standard R library).

The following functions are given to us from dplyr. We have dplyr loaded if we’ve run library(tidyverse), but I’ll include the dplyr:: first as a reminder that that’s where these functions come from.

dplyr::first(site_code)
dplyr::last(age_at_visit)
dplyr::nth(site_code, 2)
dplyr::n_distinct(site_code)

All of these functions return a single value. Try the following. What happens and why?

tumor_size * 10

age_at_visit - age_at_diagnosis

paste("Site Code:", site_code)

Because R is vectorized, operations are applied to the whole vector.

Dot, dot, dot

R has a somewhat unique addition for writing and using functions: the dot-dot-dot (...).

The ... is used in two ways:

1. To allow you to include an unknown number of values.

   paste <- function (..., sep = " ", collapse = NULL) 

   paste("a", "b", "c")
   paste("a", "b", "c", "d")

2. To allow you to pass arguments to an underlying function.

   rep <- function (x, ...)  .Primitive("rep")

   rep(1L, 4)
   rep(1L, times = 4)
   rep.int(1L, times = 4)

Before We Begin

Whenever you’re learning a new tool, for a long time you’re going to suck. It’s going to be very frustrating. But the good news is that that is typical, it’s something that happens to everyone, and it’s only temporary.

Unfortunately, there is no way to go from knowing nothing about a subject to knowing something about the subject … without going through a period of great frustration and much suckiness.

But remember, when you’re getting frustrated, that’s a good thing, it’s typical, it’s temporary. Keep pushing through and in time it will become second nature.

Hadley Wickham, UseR!2014

filter()

filter(example, site_code == "C400")
filter(example, tumor_size > 9.5)
filter(example, age_visit < max(age_dx))

Multiple arguments to filter() are combined with &:

filter(example, site_code == "C400", tumor_size > 9.5)
filter(example, site_code == "C400" & tumor_size > 9.5)

To build more complex filter combinations, use

filter(example, site_code == "C412" | "C220") # doesn't work

Error in filter_impl(.data, quo): Evaluation error: operations are possible only for numeric, logical or complex types.

filter(example, site_code == "C412" | site_code == "C220")
filter(example, site_code %in% c("C412", "C220"))

NA > 42
1 == NA
NA + 2
NA / 3
NA == NA

# Let x be Mary's age. We don't know how old she is.
x <- NA

# Let y be John's age. We don't know how old he is.
y <- NA

# Are John and Mary the same age?
x == y

[1] NA

df <- tibble(x = c(1, NA, 3))
filter(df, x > 1)
filter(df, is.na(x) | x > 1)

arrange()

To arrange, or sort, the rows according to values in a given column, use arrange().

arrange(example, site_code)
arrange(example, desc(site_code))
arrange(example, site_code, tumor_size)
arrange(example, desc(tumor_size), site_code)

select()

select(example, age_dx, tumor_size)
select(example, age_dx:tumor_size)
select(example, -(patient_id:age_visit))

select(example, starts_with("age"))
select(example, ends_with("id"))
select(example, contains("code"))

select(example, code = site_code)
rename(example, code = site_code)

mutate()

mutate(example, follow_up = age_visit - age_dx)
mutate(example, elapsed = age_visit - min(age_visit))
mutate(example, 
       tumor_size = tumor_size * 10, 
       tumor_size_mm = tumor_size *100)

mutate(example, site_code = recode(site_code, "C412" = "C999"))
mutate(example, tumor_size = if_else(site_code == "C400", tumor_size * 100, tumor_size))

summarize()

summarize(example, min(age_dx))
summarize(example, max_tumor_size = max(tumor_size))
summarize(example, tumor_size_median = median(tumor_size),tumor_size_mean = mean(tumor_size), age_visit = max(age_visit))

group_by() & summarize()

example_grp <- group_by(example, site_code)
summarize(example_grp, 
  tumor_size_mean = mean(tumor_size), 
  age_mean = mean(age_dx))

example_grp <- group_by(example, site_code, patient_id)
summarize(example_grp, 
  tumor_size_mean = mean(tumor_size), 
  age_mean = mean(age_dx))

group_by() & count()

count(example, site_code)
count(example, patient_id)

# Groups: site_code, patient_id
count(example_grp)

Your turn!

Use the pipe operator and dplyr verbs to complete the following task:

1. Use the example dataset

2. Filter out tumors smaller than 9.7

3. Calculate follow_up time as the number of years between diagnosis and the patient’s visit

4. Rename the site_code column to code.