Data wrangling
The R Bootcamp @ Erfurt
www.therbootcamp.com
@therbootcamp
Overview
In this practical you’ll practice “data wrangling” with the dplyr and tidyr packages (part of the `tidyverse collection of packages).
By the end of this practical you will know how to:
- Change column names, select specific columns
- Create new columns based on existing ones
- Select specific rows of data based on multiple criteria
- Group data and calculate summary statistics
- Combine multiple data sets through key columns
- Convert data between wide and long formats
Cheatsheet
Data wrangling with dplyr and tidyr
- Data wrangling with dplyr and tidyr Cheatsheet: https://www.rstudio.com/wp-content/uploads/2015/02/data-wrangling-cheatsheet.pdf.
Glossary
Here are the main functions you will be using in dplyr:
| Function | Description | Example |
|---|---|---|
filter() |
Select rows based on some criteria | data %>% filter(age > 40 & sex == “m”) |
arrange() |
Sort rows | data %>% arrange(date, group) |
select() |
Select columns. | data %>% select(age, sex) data %>% select(age, sex, everything()) |
rename() |
Rename columns | data %>% rename(new = old) |
mutate() |
Add new columns | data %>% mutate(height.m = height.cm / 100) |
case_when() |
Recode values of a column | data %>% mutate(sex_n = case_when(sex == 0 ~ “m”, sex == 1 ~ “f”)) |
group_by(), summarise() |
Group data and then calculate summary statistics | data %>% group_by(…) %>% summarise(…) |
left_join() |
Combine multiple data sets using a key column | data %>% left_join(data2, by = “id”) |
Here are the two functions you will be using from the tidyr package:
| Function | Description | Example |
|---|---|---|
spread() |
Convert long data to wide format - from rows to columns | data %>% gather(time, data, -id) |
gather() |
Convert wide data to long format - from columns to rows | data %>% spread(time, data) |
Examples
The following examples will take you through the steps of doing data wrangling with dplyr. Try to go through each line of code and see how it works!
# -----------------------------------------------
# Examples of using dplyr on the baselers data
# ------------------------------------------------
library(tidyverse) # Load tidyverse for dplyr and tidyr
library(skimr)
# Load baselers data
baselers <- read_csv("https://raw.githubusercontent.com/therbootcamp/baselers/master/inst/extdata/baselers.txt")
# Skim the data
skim(baselers)
baselers <- baselers %>%
# Change some names
rename(age_y = age,
swimming = rhine) %>%
# Only include people over 30
filter(age_y > 30) %>%
# Calculate some new columns
mutate(weight_lbs = weight * 2.22,
height_m = height / 100,
BMI = weight / height_m ^ 2,
# Make binary version of sex
sex_bin = case_when(
sex == "male" ~ 0,
sex == "female" ~ 1,
TRUE ~ NA_real_),
# Show when height is greater than 150
height_lt_150 = case_when(
height < 150 ~ 1,
height >= 150 ~ 0,
TRUE ~ NA_real_
) %>%
# Sort in ascending order of sex, then
# descending order of age
arrange(sex, desc(age_y)))
# Calculate grouped summary statistics
baselers_agg <- baselers %>%
group_by(sex, education) %>%
summarise(
age_mean = mean(age_y, na.rm = TRUE),
income_median = median(income, na.rm = TRUE),
N = n()
)Packages
| Package | Installation |
|---|---|
tidyverse |
install.packages("tidyverse") |
Datasets
| File | Rows | Columns |
|---|---|---|
trial_act.csv |
2139 | 27 |
trial_act_demo_fake |
2139 | 3 |
Tasks
Getting setup
A. Open your R project. It should already have the folders 0_Data and 1_Code. Make sure that the trial_act.csv and trial_act_demo_fake.csv files are in your 1_Data folder
# Done!B. Open a new R script and save it as a new file called wrangling_practical.R in the 2_Code folder. At the top of the script, using comments, write your name and the date. Then, load the set of packages for this practical with library(). Here’s how the top of your script should look:
## NAME
## DATE
## Wrangling Practical
library(XX)
library(XX)
#...C. For this practical, we’ll use the trial_act data, this is the result of a randomized clinical trial comparing the effects of different medications on adults infected with the human immunodeficiency virus. Using the following template, load the data into R and store it as a new object called trial_act.
# Load trial_act.csv from the data folder in your working directory
trial_act <- read_csv(file = "XXX/XXX")trial_act <- read_csv(file = "1_Data/trial_act.csv")Parsed with column specification:
cols(
.default = col_integer(),
wtkg = col_double()
)See spec(...) for full column specifications.D. Using the same code structure, load the trial_act_demo_fake.csv data as a new dataframe called trial_act_demo_fake
E. The trial_act data is actually a copy of a dataset from the speff2trial package called ACTG175. Look at the help menu for the ACTG175 data by running ?ACTG175 (If you become really interested in the data, you can also read an article discussing the trial here: http://www.nejm.org/doi/full/10.1056/nejm199610103351501#t=article)
# Look at documentation for ACTG175 data (contained in the speff2trial package)
?ACTG175F. Take a look at the first few rows of the datasets by printing them to the console.
# Print trial_act object
trial_actG. Use the skim() function (from the skimr package) to get more details on the datasets.
skim(trial_act)Skim summary statistics
n obs: 2139
n variables: 27
── Variable type:integer ────────────────────────────────────────────────────────────
variable missing complete n mean sd p0 p25 p50
age 0 2139 2139 35.25 8.71 12 29 34
arms 0 2139 2139 1.52 1.13 0 1 2
cd40 0 2139 2139 350.5 118.57 0 263.5 340
cd420 0 2139 2139 371.31 144.63 49 269 353
cd496 797 1342 2139 328.57 174.66 0 209.25 321
cd80 0 2139 2139 986.63 480.2 40 654 893
cd820 0 2139 2139 935.37 444.98 124 631.5 865
cens 0 2139 2139 0.24 0.43 0 0 0
days 0 2139 2139 879.1 292.27 14 727 997
drugs 0 2139 2139 0.13 0.34 0 0 0
gender 0 2139 2139 0.83 0.38 0 1 1
hemo 0 2139 2139 0.084 0.28 0 0 0
homo 0 2139 2139 0.66 0.47 0 0 1
karnof 0 2139 2139 95.45 5.9 70 90 100
offtrt 0 2139 2139 0.36 0.48 0 0 0
oprior 0 2139 2139 0.022 0.15 0 0 0
pidnum 0 2139 2139 248778.25 234237.29 10056 81446.5 190566
preanti 0 2139 2139 379.18 468.66 0 0 142
r 0 2139 2139 0.63 0.48 0 0 1
race 0 2139 2139 0.29 0.45 0 0 0
str2 0 2139 2139 0.59 0.49 0 0 1
strat 0 2139 2139 1.98 0.9 1 1 2
symptom 0 2139 2139 0.17 0.38 0 0 0
treat 0 2139 2139 0.75 0.43 0 1 1
z30 0 2139 2139 0.55 0.5 0 0 1
zprior 0 2139 2139 1 0 1 1 1
p75 p100 hist
40 70 ▁▂▇▇▃▁▁▁
3 3 ▇▁▇▁▁▇▁▇
423 1199 ▁▆▇▃▁▁▁▁
460 1119 ▂▇▇▅▂▁▁▁
440 1190 ▃▇▇▅▁▁▁▁
1207 5011 ▃▇▂▁▁▁▁▁
1146.5 6035 ▇▇▁▁▁▁▁▁
0 1 ▇▁▁▁▁▁▁▂
1091 1231 ▁▂▂▂▂▂▇▇
0 1 ▇▁▁▁▁▁▁▁
1 1 ▂▁▁▁▁▁▁▇
0 1 ▇▁▁▁▁▁▁▁
1 1 ▅▁▁▁▁▁▁▇
100 100 ▁▁▁▁▁▅▁▇
1 1 ▇▁▁▁▁▁▁▅
0 1 ▇▁▁▁▁▁▁▁
280277 990077 ▇▇▅▁▁▁▁▂
739.5 2851 ▇▂▂▁▁▁▁▁
1 1 ▅▁▁▁▁▁▁▇
1 1 ▇▁▁▁▁▁▁▃
1 1 ▆▁▁▁▁▁▁▇
3 3 ▇▁▁▃▁▁▁▇
0 1 ▇▁▁▁▁▁▁▂
1 1 ▂▁▁▁▁▁▁▇
1 1 ▆▁▁▁▁▁▁▇
1 1 ▁▁▁▇▁▁▁▁
── Variable type:numeric ────────────────────────────────────────────────────────────
variable missing complete n mean sd p0 p25 p50 p75 p100
wtkg 0 2139 2139 75.13 13.26 31 66.68 74.39 82.56 159.94
hist
▁▂▇▅▁▁▁▁Notes
In this practical you’ll be doing lots of sequential operations on the same data. Whenever you can, try to keep the pipe %>% going to connect each task to the one before it. See below for an example
# Version A) Try to convert this...
baselers <- baselers %>% rename(age_y = age)
baselers <- baselers %>% rename(swimming = rhine)
baselers <- baselers %>% filter(age_y > 30)
# Version B) To this version where you keep the pipe going from the beginning!
baselers <- baselers %>%
rename(age_y = age,
swimming = rhine) %>%
filter(age_y > 30)Change column names with rename()
Let’s change some of the column names in
trial_act. Usingrename(), change the column namewtkgtoweight_kg(to specify that weight is in kilograms) Be sure to assign the result back totrial_actto change it!Change the column name
agetoage_y(to specify that age is in years).
trial_act <- trial_act %>%
rename(weight_kg = wtkg,
age_y = age)Select columns with select()
Using the
select()function, create a new dataframe calledCD4_widethat only contains the columnspidnum,arms,cd40,cd420, andcd496. Thecd40,cd420, andcd496columns show patient’s CD4 T cell counts at baseline, 20 weeks, and 96 weeks. Print the result to make sure it worked!Did you know you can easily select all columns that start with specific characters using
starts_with()? Try adapting the following code to get the same result you got before.
CD4_wide <- trial_act %>%
select(pidnum, arms, starts_with("XXX"))CD4_wide <- trial_act %>%
select(pidnum, arms, starts_with("cd"))Add new columns with mutate()
- Using the
mutate()function, add the following two new columns totrial_act. Try combining these into one call to the mutate() function!agem: Each patient’s age in months instead of years (Hint: Just multiplyage_yby 12!).weight_lb: Weight in lbs instead of kilograms. You can do this by multiplyingweight_kgby 2.2.cd_change_20: Change in CD4 T cell count from baseline to 20 weeks. You can do this by takingcd420 - cd40cd_change_960: Change in CD4 T cell count from baseline to 96 weeks. You can do this by takingcd496 - cd40
trial_act <- trial_act %>%
mutate(agem = age_y * 12)mutate(case_when())
- Create a new column
gender_charthat shows gender as a character string. To do this, use a combination ofmutate()andcase_when. The original gender data is stored in thegendercolumn, where 0 = “female” and 1 = “male”.
trial_act <- trial_act %>%
mutate(
gender_char = case_when(
gender == 0 ~ "female",
gender == 1 ~ "male"
)
)- Create a new column
over50that is 1 when patients are older than 50, and 0 when they are younger than or equal to 50 (hint: Use logical comparisons > and <=)
trial_act <- trial_act %>%
mutate(
over50 = case_when(
age_y > 50 ~ 1,
age_y <= 50 ~ 0
)
)- If you haven’t already, put the code for your previous questions in one call to
mutate(). That is, in one block of code, createagem,weight_lb,cd_change_20,cd_change_960,gender_charandover50using themutate()function only once. Here’s how your code should look:
trial_act <- trial_act %>%
mutate(
agem = XXX,
weight_lb = XXX,
cd_change_20 = XXX,
cd_change_960 = XXX,
gender_char = case_when(XXX),
over50 = case_when(XXX)
)trial_act <- trial_act %>%
mutate(
agem = agem = age_y * 12,
weight_lb = weight_kg * 2.2,
cd_change_20 = cd420 - cd40,
cd_change_960 = cd496 - cd40,
gender_char = case_when(
gender == 0 ~ "female",
gender == 1 ~ "male"
),
over50 = case_when(
age_y > 50 ~ 1,
age_y <= 50 ~ 0
)
)Arrange rows with arrange()
- Using the
arrange()function, arrange thetrial_actdata in ascending order ofage_y(from lowest to highest). After you do, look the data to make sure it worked!
trial_act <- trial_act %>%
arrange(age_y)
trial_act# A tibble: 2,139 x 30
pidnum age_y weight_kg hemo homo drugs karnof oprior z30 zprior
<int> <int> <dbl> <int> <int> <int> <int> <int> <int> <int>
1 940533 12 41.4 1 0 0 100 0 0 1
2 950037 12 53.1 1 0 0 100 0 1 1
3 950056 12 31 1 0 0 100 0 1 1
4 910034 13 32.7 1 0 0 100 0 1 1
5 940534 13 62.9 1 0 0 100 0 0 1
6 960014 13 48.5 1 0 0 100 0 0 1
7 310767 14 65 1 0 0 100 0 1 1
8 920050 14 54.2 1 0 0 100 0 1 1
9 940544 14 41.1 1 0 0 100 0 0 1
10 950061 14 64.3 1 0 0 100 0 1 1
# ... with 2,129 more rows, and 20 more variables: preanti <int>,
# race <int>, gender <int>, str2 <int>, strat <int>, symptom <int>,
# treat <int>, offtrt <int>, cd40 <int>, cd420 <int>, cd496 <int>,
# r <int>, cd80 <int>, cd820 <int>, cens <int>, days <int>, arms <int>,
# agem <dbl>, gender_char <chr>, over50 <dbl>- Now arrange the data in descending order of
age_y(from highest to lowest). After, look the data to make sure it worked. To arrange data in descending order, just includedesc()around the variable. E.g.;data %>% arrrange(desc(height))
trial_act <- trial_act %>%
arrange(desc(age_y))
trial_act# A tibble: 2,139 x 30
pidnum age_y weight_kg hemo homo drugs karnof oprior z30 zprior
<int> <int> <dbl> <int> <int> <int> <int> <int> <int> <int>
1 11438 70 73.9 0 1 0 100 0 1 1
2 211360 70 63.1 0 1 0 90 0 0 1
3 211284 69 81.6 0 1 0 100 0 0 1
4 50580 68 90.5 0 1 1 100 0 1 1
5 81127 68 70.8 0 1 0 90 0 1 1
6 10924 67 71 0 1 0 100 0 0 1
7 241150 67 82.1 0 1 0 90 0 0 1
8 50662 66 84.4 0 1 0 100 0 0 1
9 11987 65 77.2 0 1 0 90 0 0 1
10 140797 65 60.5 0 1 0 90 0 0 1
# ... with 2,129 more rows, and 20 more variables: preanti <int>,
# race <int>, gender <int>, str2 <int>, strat <int>, symptom <int>,
# treat <int>, offtrt <int>, cd40 <int>, cd420 <int>, cd496 <int>,
# r <int>, cd80 <int>, cd820 <int>, cens <int>, days <int>, arms <int>,
# agem <dbl>, gender_char <chr>, over50 <dbl>- You can sort the rows of dataframes with multiple columns by including many arguments to
arrange(). Now sort the data by arms (arms) and then age (age_y).
trial_act <- trial_act %>%
arrange(arms, age_y)
trial_act# A tibble: 2,139 x 30
pidnum age_y weight_kg hemo homo drugs karnof oprior z30 zprior
<int> <int> <dbl> <int> <int> <int> <int> <int> <int> <int>
1 960014 13 48.5 1 0 0 100 0 0 1
2 960031 14 48.3 1 0 0 100 0 0 1
3 990071 14 60 1 0 0 100 0 0 1
4 980042 16 63 1 0 0 100 0 1 1
5 171040 17 51.3 0 0 0 90 0 0 1
6 990026 17 103. 1 0 0 100 0 1 1
7 310234 18 57.3 1 0 0 100 0 1 1
8 940519 18 56.8 1 0 0 100 0 1 1
9 211314 19 50.8 0 0 0 90 0 0 1
10 340767 19 74.8 0 1 0 100 0 0 1
# ... with 2,129 more rows, and 20 more variables: preanti <int>,
# race <int>, gender <int>, str2 <int>, strat <int>, symptom <int>,
# treat <int>, offtrt <int>, cd40 <int>, cd420 <int>, cd496 <int>,
# r <int>, cd80 <int>, cd820 <int>, cens <int>, days <int>, arms <int>,
# agem <dbl>, gender_char <chr>, over50 <dbl>Filter specific rows with filter()
- Using the
filter()function, create a new dataframe calledtrial_act_mthat only contains data from males (gender_char == "male")
trial_act_B <- trial_act %>%
filter(gender == 1)- A colleague of yours named Tracy wants a datafame only containing data from females over the age of 40. Create this dataframe with
filter()and call ittrial_act_Tracy
trial_act_C <- trial_act %>%
filter(age_y > 40 & gender == 0)Combine dataframes with left_join()
- The
trial_act_demo_fake.csvfile contains additional (fictional) demographic data about the patients, namely the number of days of exercise they get per week, and their highest level of education. Use theleft_join()function to combine thetrial_actandtrial_act_demo_fakedatasets, set thebyargument to the name of the column that is common in both data sets. This will be the key! Assign the result totrial_act. When you are done, look at thetrial_actdataframe to make sure your code worked!
trial_act_demo_fake <- read_csv("1_Data/trial_act_demo_fake.csv")Parsed with column specification:
cols(
pidnum = col_integer(),
exercise = col_integer(),
education = col_character()
)trial_act <- trial_act %>%
left_join(trial_act_demo_fake, by = "pidnum")- Using your new
trial_actdataframe, which should contain the exercise data, calculate the mean number of days of exercise that patients reported.
mean(trial_act$exercise)[1] 2.203834Calculate grouped statistics with group_by() and summarise()
- In this code we’ll calculate summary statistics for each of the trial arms. Start with the
trial_actdataframe. Then, group the data byarms. Then, for each arm, calculate the mean participant age (in years) as a new column calledage_mean. Also, usingN = n(), calculate the number of cases for each group. Assign the result to a new object calledtrial_arm.
trial_act %>%
group_by(arms) %>%
summarise(
age_mean = mean(agey)
)- Now adjust your previous code to calculate the standard deviation of age in addition to the mean.
trial_act %>%
group_by(arms) %>%
summarise(
age_mean = mean(age_y),
karnof_median = median(karnof)
)# A tibble: 4 x 3
arms age_mean karnof_median
<int> <dbl> <dbl>
1 0 35.2 100
2 1 35.2 100
3 2 35.4 100
4 3 35.1 100- Add code to calculate the median number of days until the first major negative event (
days) for each arm.
trial_act %>%
group_by(arms) %>%
summarise(
age_mean = mean(age_y),
karnof_median = median(karnof),
days_median = median(days)
)# A tibble: 4 x 4
arms age_mean karnof_median days_median
<int> <dbl> <dbl> <dbl>
1 0 35.2 100 934.
2 1 35.2 100 1014
3 2 35.4 100 1012
4 3 35.1 100 1000 - Your code so far only creates groups based on
arm, now adjust the code so it creates groups based ongender_charonly (that is, forget about the trial arm). Keep calculating all of the same summary statistics as you did before. Assign the result to a new dataframe calledtrial_gender
trial_gender <- trial_act %>%
group_by(gender_char) %>%
summarise(
age_mean = mean(age_y),
karnof_median = median(karnof),
days_median = median(days)
)
trial_gender# A tibble: 2 x 4
gender_char age_mean karnof_median days_median
<chr> <dbl> <dbl> <dbl>
1 female 34.3 100 982
2 male 35.4 100 1002- Now group the data by both
armandgender_charand calculate the same summary statistics! Assign the result to a new object calledtrial_arm_gen
trial_gender <- trial_act %>%
group_by(gender_char, arms) %>%
summarise(
age_mean = mean(age_y),
karnof_median = median(karnof),
days_median = median(days)
)
trial_gender# A tibble: 8 x 5
# Groups: gender_char [?]
gender_char arms age_mean karnof_median days_median
<chr> <int> <dbl> <dbl> <dbl>
1 female 0 33.6 100 936.
2 female 1 33.5 100 995
3 female 2 35.0 100 1004
4 female 3 35.2 100 974
5 male 0 35.6 100 932.
6 male 1 35.6 100 1024
7 male 2 35.5 100 1013
8 male 3 35.1 100 1010 Reshaping with gather() and spread()
- Remember the
CD4_widedataframe you created before? Currently it is in the wide format, where key data (different CD4 T cell counts) are in different columns. Now we will try to convert it to a long format. Our goal is to get the data in the ‘long’ format. Using thespread()function, create a new dataframe calledCD4_longthat shows the data in the ‘long’ format. To do this, use the following template. Set the grouping column totimeand the new data column tovalue.
CD4_long <- CD4_wide %>%
gather(XX, # New grouping column
XX, # New data column
-pidnum, -arms) # Names of columns to replicateCD4_long <- CD4_wide %>%
gather(time, # New grouping column
value, # New data column
-pidnum, -arms) # Names of columns to replicate- Now that your data are in the wide format, it should be easy to calculate grouped summary statistics! For each time point and trial arm, calculate the mean CD4 T cell count using
group_by()andsummarise().
CD4_long %>%
group_by(time, arms) %>%
summarise(
cd4_mean = mean(value, na.rm = TRUE)
)# A tibble: 20 x 3
# Groups: time [?]
time arms cd4_mean
<chr> <int> <dbl>
1 cd40 0 353.
2 cd40 1 349.
3 cd40 2 353.
4 cd40 3 347.
5 cd420 0 336.
6 cd420 1 403.
7 cd420 2 372.
8 cd420 3 374.
9 cd496 0 288.
10 cd496 1 341.
11 cd496 2 355.
12 cd496 3 329.
13 cd80 0 987.
14 cd80 1 1004.
15 cd80 2 984.
16 cd80 3 972.
17 cd820 0 928.
18 cd820 1 968.
19 cd820 2 902.
20 cd820 3 943.- Now it’s time to practice moving data from the long to the wide format. Using the following template, use the
spread()function to convertCD4_longback the wide format. Assign the result to a new object calledCD4_wide_2. It should look exactly likeCD4_wide!
CD4_wide_2 <- CD4_long %>%
spread(time, # old group column
value) # old target column
CD4_wide_2Play around with “Scoped” functions
Many common dplyr functions like mutate() and summarise() have ‘scoped’ versions with suffixes like _if and _all. that allow you do some pretty cool stuff easily (look at the help menu with ?scoped for details). Try running the following chunk with summarise_if() and see what happens:
baselers <- read_csv("1_Data/baselers.csv")Parsed with column specification:
cols(
.default = col_integer(),
sex = col_character(),
height = col_double(),
weight = col_double(),
income = col_double(),
education = col_character(),
confession = col_character(),
food = col_double(),
fasnacht = col_character(),
eyecor = col_character()
)See spec(...) for full column specifications.# See how summerise_if() works!
baselers %>%
group_by(sex) %>%
summarise_if(is.numeric, mean)# A tibble: 2 x 16
sex id age height weight income children happiness fitness food
<chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 female 5016. 45.4 164. NA NA 1.81 6.91 5.12 NA
2 male 4985. 43.8 178. NA NA 1.82 6.90 5.13 NA
# ... with 6 more variables: alcohol <dbl>, tattoos <dbl>, rhine <dbl>,
# datause <dbl>, consultations <dbl>, hiking <dbl>- Now, in the
trial_actdataset, group the data byarmsand calculate the mean of all numeric columns usingsummerise_if()
trial_act %>%
group_by(arms) %>%
summarise_if(is.numeric, mean, na.rm = TRUE)# A tibble: 4 x 30
arms pidnum age_y weight_kg hemo homo drugs karnof oprior z30
<int> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 0 252884. 35.2 76.1 0.0789 0.641 0.118 95.4 0.0301 0.547
2 1 232609. 35.2 74.9 0.0824 0.663 0.140 95.5 0.0172 0.552
3 2 257593. 35.4 74.7 0.0878 0.664 0.145 95.7 0.0248 0.561
4 3 251696. 35.1 74.9 0.0873 0.676 0.123 95.1 0.0160 0.542
# ... with 20 more variables: zprior <dbl>, preanti <dbl>, race <dbl>,
# gender <dbl>, str2 <dbl>, strat <dbl>, symptom <dbl>, treat <dbl>,
# offtrt <dbl>, cd40 <dbl>, cd420 <dbl>, cd496 <dbl>, r <dbl>,
# cd80 <dbl>, cd820 <dbl>, cens <dbl>, days <dbl>, agem <dbl>,
# over50 <dbl>, exercise <dbl>Here’s another scoped function in action mutate_if() in action:
# use mutate_if() to round all numeric variables to 2 digits
baselers %>%
mutate_if(is.numeric, round, 2)- Using
mutate_if(), round all of your results from the previous question to 0 decimal places (to the nearest integer)
trial_act %>%
group_by(arms) %>%
summarise_if(is.numeric, mean, na.rm = TRUE) %>%
mutate_if(is.numeric, round, 0)# A tibble: 4 x 30
arms pidnum age_y weight_kg hemo homo drugs karnof oprior z30
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 0 252884 35 76 0 1 0 95 0 1
2 1 232609 35 75 0 1 0 96 0 1
3 2 257593 35 75 0 1 0 96 0 1
4 3 251696 35 75 0 1 0 95 0 1
# ... with 20 more variables: zprior <dbl>, preanti <dbl>, race <dbl>,
# gender <dbl>, str2 <dbl>, strat <dbl>, symptom <dbl>, treat <dbl>,
# offtrt <dbl>, cd40 <dbl>, cd420 <dbl>, cd496 <dbl>, r <dbl>,
# cd80 <dbl>, cd820 <dbl>, cens <dbl>, days <dbl>, agem <dbl>,
# over50 <dbl>, exercise <dbl>More practice
- Now let’s check the major differences between the treatment arms. For each arm, calculate the following:
- Mean days until a a major negative event (
days) - Mean CD4 T cell count at baseline. (
cd40) - Mean CD4 T cell count at 20 weeks. (
cd420) - Mean CD4 T cell count at 96 weeks. (
cd496) - Mean change in CD4 T cell count between baseline and 96 weeks
- Number of patients in each arm
trial_act %>%
group_by(arms) %>%
summarise(
days_mean = mean(days),
cd4_bl = mean(cd40),
cd4_20 = mean(cd420),
cd4_96 = mean(cd496, na.rm = TRUE),
cd4_change = mean(cd496 - cd40, na.rm = TRUE),
N = n()
)# A tibble: 4 x 7
arms days_mean cd4_bl cd4_20 cd4_96 cd4_change N
<int> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 0 801. 353. 336. 288. -77.5 532
2 1 916. 349. 403. 341. -6.90 522
3 2 906. 353. 372. 355. -4.36 524
4 3 893. 347. 374. 329. -19.4 561- Repeat the previous analysis, but before you do the grouping and summary statistics, create a new variable called
arms_charthat shows the values ofarmsas characters that reflect what the values actually represent (hint: usemutate()andcase_when()). For example, looking at the help file?ACTG175, I can see that the treatment arm of 0 is “zidovudine”. I might call this arm"Z". Do this in the all in the same chunk of code.
trial_act %>%
mutate(
arms_char = case_when(
arms == 0 ~ "Z",
arms == 1 ~ "ZD",
arms == 2 ~ "ZZ",
arms == 3 ~ "D"
)
) %>%
group_by(arms_char) %>%
summarise(
days_mean = mean(days),
cd4_bl = mean(cd40),
cd4_20 = mean(cd420),
cd4_96 = mean(cd496, na.rm = TRUE),
cd4_change = mean(cd496 - cd40, na.rm = TRUE)
)# A tibble: 4 x 6
arms_char days_mean cd4_bl cd4_20 cd4_96 cd4_change
<chr> <dbl> <dbl> <dbl> <dbl> <dbl>
1 D 893. 347. 374. 329. -19.4
2 Z 801. 353. 336. 288. -77.5
3 ZD 916. 349. 403. 341. -6.90
4 ZZ 906. 353. 372. 355. -4.36- Create the following dataframe that shows patient’s mean CD8 T cell count (from columns
cd80andcd820), where the data are grouped by time and trial arm. (Hint: use the following functions in order: select(), gather(), mutate(), group_by(), summarise())
trial_act %>%
mutate(
arms_char = case_when(
arms == 0 ~ "Z",
arms == 1 ~ "ZD",
arms == 2 ~ "ZZ",
arms == 3 ~ "D"
)
) %>%
select(pidnum, arms_char, starts_with("cd8")) %>%
gather(time, measure, -pidnum, -arms_char) %>%
mutate(time = case_when(time == "cd80" ~ "baseline",
time == "cd820" ~ "week 20")) %>%
group_by(time, arms_char) %>%
summarise(N = n(),
cd8_mean = mean(measure),
cd8_median = median(measure))# A tibble: 8 x 5
# Groups: time [?]
time arms_char N cd8_mean cd8_median
<chr> <chr> <int> <dbl> <dbl>
1 baseline D 561 972. 890
2 baseline Z 532 987. 881
3 baseline ZD 522 1004. 917
4 baseline ZZ 524 984. 898.
5 week 20 D 561 943. 871
6 week 20 Z 532 928. 818
7 week 20 ZD 522 968. 903
8 week 20 ZZ 524 902. 862