Session 8

Working with Strings

Question: You have a vector that contains two strings.

library(stringr)
txt <- c("pfx-003-sfx", "pfx-987-sfx")

Match the function that was used to the output that it created.

1. str_detect(txt, "003")
2. str_extract(txt, "\\d+")
3. str_remove(txt, "-sfx")
4. str_c(txt, collapse = " and ")
5. str_replace(txt, "pfx", "prefix")

1. c("prefix-003-sfx", "prefix-987-sfx")
2. c("003", "987")
3. c("pfx-003", "pfx-987")
4. c(TRUE, FALSE)
5. "pfx-003-sfx and pfx-987-sfx"

Question: The following ICD10 codes describe a selection of ICD codes for “Malignant melanomas of the skin”.

1. Describe in words all of the patterns you can find in the ICD code and its description.

2. How can you extract the specific location referred to in the description text?

3. How could you find all patients with melonoma located on the eyelid? On the ear?

4. Write a regular expression that finds ICD codes indicate melanoma on the right side of the body.

Another way to do things

R is about 25 years old, or 42 if you count its “first version” S.

dplyr, on the other hand, started in October of 2012 but the version most consistent with the things that we’ve learned wasn’t released until October 2014 – or 4.1 years ago.

This means that when you run into issues and search around online in places like StackOverflow, you’ll often run into R code written in base R. It’s also useful to know a bit about base R functions when you need to do more programming-style work, as opposed to the data analysis workflows we’ve been learning.

This section is intended to give a very quick overview of several parts of base R that we haven’t talked about much, primarily to help you translate code you find online into dplyr-based code that you’re used to.

We’ll use the example data frame from Session 3 for this section.

> x

# 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     

Selecting Elements of a Vector

First, let’s say that you have a vector.

v <- c(10, 14, 6)

You can get a particular element of the vector using [ ].

v[2]

## [1] 14

Sidenote: does R start counting elements from 0 or from 1?

You can extract more than one element by using a vector.

v[c(1, 3)]

## [1] 10  6

You can discard elements with negative indices.

v[-2]

## [1] 10  6

And you can use TRUE and FALSE to discard elements that are in the same position as the FALSE.

v != 14

## [1]  TRUE FALSE  TRUE

v[v != 14]

## [1] 10  6

v[c(TRUE, FALSE, TRUE)]

## [1] 10  6

Suppose this vector is actually a count of the number of fruit that we have. If the vector has names, we can use the names to select specific elements.

names(v) <- c("kiwi", "lime", "mango")
v["mango"]

## mango 
##     6

v[c("kiwi", "lime")]

## kiwi lime 
##   10   14

Suppose instead of a vector, we have a list.

l <- list(
  orange = 25,
  plum = 13,
  quince = 30
)

Yes, it’s a real fruit

Lists are slightly more tricky. Using one set of [ ] gives you back another list with the elements in the brackets, or you can get a specific element with double brackets [[ ]].

l["orange"]

## $orange
## [1] 25

l[c("orange", "plum")]

## $orange
## [1] 25
## 
## $plum
## [1] 13

l[["quince"]]

## [1] 30

The double brackets can only give you one entry; you’ll get an error if you try to get more than one.

l[[c("orange", "plum")]]

## Error in l[[c("orange", "plum")]]: subscript out of bounds

Did you notice the $ in the output? That’s because the dollar sign is yet another subsetting operator and it works exactly the same as [[ ]].

l$plum

## [1] 13

l[["plum"]]

## [1] 13

Why are there two three ways to subset lists? First: why does there need to be a difference between [ ] and [[ ]]?

l$other <- c("rambutan", "starfruit", "tomato")
l

## $orange
## [1] 25
## 
## $plum
## [1] 13
## 
## $quince
## [1] 30
## 
## $other
## [1] "rambutan"  "starfruit" "tomato"

Yup, it really is, too…

l[c("orange", "other")]

## $orange
## [1] 25
## 
## $other
## [1] "rambutan"  "starfruit" "tomato"

l[["orange"]]

## [1] 25

l[["other"]]

## [1] "rambutan"  "starfruit" "tomato"

c(l$other, l$orange)

## [1] "rambutan"  "starfruit" "tomato"    "25"

Subsetting Columns and Rows

The syntax for data frames (or tibbles) is very similar. The only difference is that the single brackets [ ] take two arguments – rows and columns – and the double brackets [[ ]] (or $) return columns.

x

## # 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

x[1, 2]

## # A tibble: 1 x 1
##   age_dx
##    <dbl>
## 1     54

x[c(1, 4), c(2, 3)]

## # A tibble: 2 x 2
##   age_dx age_visit
##    <dbl>     <int>
## 1     54        54
## 2     54        57

x[1:2, c("age_dx", "age_visit")]

## # A tibble: 2 x 2
##   age_dx age_visit
##    <dbl>     <int>
## 1     54        54
## 2     54        55

x$patient_id

## [1] 5554321 5554321 5554321 5554321 5554321

x[["age_dx"]]

## [1] 54 54 54 54 54

What if you just want to select certain rows or certain columns?

x[c(1, 4), ]

## # A tibble: 2 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        57        9.9 C220

x[, 1:2]

## # A tibble: 5 x 2
##   patient_id age_dx
##        <dbl>  <dbl>
## 1    5554321     54
## 2    5554321     54
## 3    5554321     54
## 4    5554321     54
## 5    5554321     54

What does this code do?

x[x$age_visit > x$age_dx, ]$tumor_size

## [1]  9.5  9.7  9.9 10.1

# Base R
x[, c("age_dx", "age_visit")]

x %>% select(age_dx, age_visit)

## # A tibble: 5 x 2
##   age_dx age_visit
##    <dbl>     <int>
## 1     54        54
## 2     54        55
## 3     54        56
## 4     54        57
## 5     54        58

# Base R
x[c(1, 4), ]

x %>% slice(c(1, 4))

## # A tibble: 2 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        57        9.9 C220

# Base R
x[1, 2]

x %>% 
  slice(1) %>% 
  select(2)

## # A tibble: 1 x 1
##   age_dx
##    <dbl>
## 1     54

# Base R
x[x$age_visit > x$age_dx, "tumor_size"]

x %>% 
  filter(age_visit > age_dx) %>% 
  select(tumor_size)

## # A tibble: 4 x 1
##   tumor_size
##        <dbl>
## 1        9.5
## 2        9.7
## 3        9.9
## 4       10.1

# Base R
x[x$age_visit > x$age_dx, ]$tumor_size

x %>% 
  filter(age_visit > age_dx) %>% 
  pull(tumor_size)

## [1]  9.5  9.7  9.9 10.1

WTF: What the Factor?

Imagine you have a variable that records months:

m <- c("Dec", "Oct", "Sep", "May")

Using a string for this variable has two problems:

1. There are only 12 possible values. What if you make a mistake?

   m2 <- c("Dev", "Oct", "Seo", "May")

2. You can’t sort in a useful way.

   sort(m)

   ## [1] "Dec" "May" "Oct" "Sep"

A factor solves both of these problems:

1. It adds the concept of a categorical level

   month_levels <- c(
     "Jan", "Feb", "Mar", "Apr", "May", "Jun", 
     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
   )
   m <- factor(m, levels = month_levels)
   m

   ## [1] Dec Oct Sep May
   ## Levels: Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

2. that have an explicit order

   sort(m)

   ## [1] May Sep Oct Dec
   ## Levels: Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

If you try to create a factor with a typo, you’ll get a missing value

factor(m2, levels = month_levels)

## [1] <NA> Oct  <NA> May 
## Levels: Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

library(tidyverse)

# or
library(forcats)

Bi-Weekly Workshops

Every other week, starting the week of September 17, 2018. What time works for you?

http://whenisgood.net/gerkelab/cds-r-follow-up

Potential Topics

1. Creating data-driven reports with R Markdown.

2. Visualizing data with ggplot2

3. Writing functions for repeatable analysis

4. Using version control software to track changes

5. Accessing data stored in databases with dbplyr

Stay in Touch!