Learning Objectives
- Understand the purpose of the
dplyr
andtidyr
packages.- Select certain columns in a data frame with the
dplyr
functionselect
.- Select certain rows in a data frame according to filtering conditions with the
dplyr
functionfilter
.- Link the output of one
dplyr
function to the input of another function with the ‘pipe’ operator%>%
.- Add new columns to a data frame that are functions of existing columns with
mutate
.- Understand the split-apply-combine concept for data analysis.
- Use
summarize
,group_by
, andtally
to split a data frame into groups of observations, apply a summary statistics for each group, and then combine the results.- Understand the concept of a wide and a long table format and for which purpose those formats are useful.
- Understand what key-value pairs are.
- Reshape a data frame from long to wide format and back with the
spread
andgather
commands from thetidyr
package.- Export a data frame to a .csv file.
dplyr
and tidyr
Bracket subsetting is handy, but it can be cumbersome and difficult to read, especially for complicated operations. Enter dplyr
. dplyr
is a package for making tabular data manipulation easier. It pairs nicely with tidyr
which enables you to swiftly convert between different data formats for plotting and analysis.
Packages in R are basically sets of additional functions that let you do more stuff. The functions we’ve been using so far, like str()
or data.frame()
, come built into R; packages give you access to more of them. Before you use a package for the first time you need to install it on your machine, and then you should import it in every subsequent R session when you need it. You should already have installed the tidyverse
package. This is an “umbrella-package” that installs several packages useful for data analysis which work together well such as tidyr
, dplyr
, ggplot2
, etc. To load the package type:
library("tidyverse") ## load the tidyverse packages, incl. dplyr
#> Warning: Installed Rcpp (0.12.12) different from Rcpp used to build dplyr (0.12.11).
#> Please reinstall dplyr to avoid random crashes or undefined behavior.
dplyr
and tidyr
?The package dplyr
provides easy tools for the most common data manipulation tasks. It is built to work directly with data frames, with many common tasks optimized by being written in a compiled language (C++). An additional feature is the ability to work directly with data stored in an external database. The benefits of doing this are that the data can be managed natively in a relational database, queries can be conducted on that database, and only the results of the query are returned.
This addresses a common problem with R in that all operations are conducted in-memory and thus the amount of data you can work with is limited by available memory. The database connections essentially remove that limitation in that you can have a database of many 100s GB, conduct queries on it directly, and pull back into R only what you need for analysis.
The package tidyr
addresses the common problem of wanting to reshape your data for plotting and use by different R functions. Sometimes we want data sets where we have one row per measurement. Sometimes we want a data frame where each measurement type has its own column, and rows are instead more aggregated groups - like plots or aquaria. Moving back and forth between these formats is nontrivial, and tidyr
gives you tools for this and more sophisticated data manipulation.
To learn more about dplyr
and tidyr
after the workshop, you may want to check out this handy data wrangling cheatsheet.
We’re going to learn some of the most common dplyr
functions: select()
, filter()
, mutate()
, group_by()
, and summarize()
. To select columns of a data frame, use select()
. The first argument to this function is the data frame (surveys
), and the subsequent arguments are the columns to keep.
select(surveys, plot_id, species_id, weight)
To choose rows based on a specific criteria, use filter()
:
filter(surveys, year == 1995)
But what if you wanted to select and filter at the same time? There are three ways to do this: use intermediate steps, nested functions, or pipes.
With intermediate steps, you essentially create a temporary data frame and use that as input to the next function. This can clutter up your workspace with lots of objects. You can also nest functions (i.e. one function inside of another). This is handy, but can be difficult to read if too many functions are nested as things are evaluated from the inside out.
The last option, pipes, are a fairly recent addition to R. Pipes let you take the output of one function and send it directly to the next, which is useful when you need to do many things to the same dataset. Pipes in R look like %>%
and are made available via the magrittr
package, installed automatically with dplyr
. If you use RStudio, you can type the pipe with Ctrl + Shift + M if you have a PC or Cmd + Shift + C if you have a Mac.
surveys %>%
filter(weight < 5) %>%
select(species_id, sex, weight)
In the above, we use the pipe to send the surveys
dataset first through filter()
to keep rows where weight
is less than 5, then through select()
to keep only the species_id
, sex
, and weight
columns. Since %>%
takes the object on its left and passes it as the first argument to the function on its right, we don’t need to explicitly include it as an argument to the filter()
and select()
functions anymore.
If we wanted to create a new object with this smaller version of the data, we could do so by assigning it a new name:
surveys_sml <- surveys %>%
filter(weight < 5) %>%
select(species_id, sex, weight)
surveys_sml
Note that the final data frame is the leftmost part of this expression.
Challenge
Using pipes, subset the
survey
data to include individuals collected before 1995 and retain only the columnsyear
,sex
, andweight
.
Frequently you’ll want to create new columns based on the values in existing columns, for example to do unit conversions, or find the ratio of values in two columns. For this we’ll use mutate()
.
To create a new column of weight in kg:
surveys %>%
mutate(weight_kg = weight / 1000)
You can also create a second new column based on the first new column within the same call of mutate()
:
surveys %>%
mutate(weight_kg = weight / 1000,
weight_kg2 = weight_kg * 2)
If this runs off your screen and you just want to see the first few rows, you can use a pipe to view the head()
of the data. (Pipes work with non-dplyr
functions, too, as long as the dplyr
or magrittr
package is loaded).
surveys %>%
mutate(weight_kg = weight / 1000) %>%
head
Note that we don’t include parentheses at the end of our call to head()
above. When piping into a function with no additional arguments, you can call the function with or without parentheses (e.g. head
or head()
).
The first few rows of the output are full of NA
s, so if we wanted to remove those we could insert a filter()
in the chain:
surveys %>%
filter(!is.na(weight)) %>%
mutate(weight_kg = weight / 1000) %>%
head
is.na()
is a function that determines whether something is an NA
. The !
symbol negates the result, so we’re asking for everything that is not an NA
.
Challenge
Create a new data frame from the
surveys
data that meets the following criteria: contains only thespecies_id
column and a new column calledhindfoot_half
containing values that are half thehindfoot_length
values. In thishindfoot_half
column, there are noNA
s and all values are less than 30.Hint: think about how the commands should be ordered to produce this data frame!
Many data analysis tasks can be approached using the split-apply-combine paradigm: split the data into groups, apply some analysis to each group, and then combine the results. dplyr
makes this very easy through the use of the group_by()
function.
summarize()
functiongroup_by()
is often used together with summarize()
, which collapses each group into a single-row summary of that group. group_by()
takes as arguments the column names that contain the categorical variables for which you want to calculate the summary statistics. So to view the mean weight
by sex:
surveys %>%
group_by(sex) %>%
summarize(mean_weight = mean(weight, na.rm = TRUE))
You may also have noticed that the output from these calls doesn’t run off the screen anymore. That’s because dplyr
has changed our data.frame
object to an object of class tbl_df
, also known as a “tibble”. Tibble’s data structure is very similar to a data frame. For our purposes the only differences are that, (1) in addition to displaying the data type of each column under its name, it only prints the first few rows of data and only as many columns as fit on one screen, (2) columns of class character
are never converted into factors.
You can also group by multiple columns:
surveys %>%
group_by(sex, species_id) %>%
summarize(mean_weight = mean(weight, na.rm = TRUE))
When grouping both by sex
and species_id
, the first rows are for individuals that escaped before their sex could be determined and weighted. You may notice that the last column does not contain NA
but NaN
(which refers to “Not a Number”). To avoid this, we can remove the missing values for weight before we attempt to calculate the summary statistics on weight. Because the missing values are removed, we can omit na.rm = TRUE
when computing the mean:
surveys %>%
filter(!is.na(weight)) %>%
group_by(sex, species_id) %>%
summarize(mean_weight = mean(weight))
Here, again, the output from these calls doesn’t run off the screen anymore. Recall that dplyr
has changed our object fromdata.frame
to tbl_df
. If you want to display more data, you can use the print()
function at the end of your chain with the argument n
specifying the number of rows to display:
surveys %>%
filter(!is.na(weight)) %>%
group_by(sex, species_id) %>%
summarize(mean_weight = mean(weight)) %>%
print(n = 15)
Once the data are grouped, you can also summarize multiple variables at the same time (and not necessarily on the same variable). For instance, we could add a column indicating the minimum weight for each species for each sex:
surveys %>%
filter(!is.na(weight)) %>%
group_by(sex, species_id) %>%
summarize(mean_weight = mean(weight),
min_weight = min(weight))
When working with data, it is also common to want to know the number of observations found for each factor or combination of factors. For this, dplyr
provides tally()
. For example, if we wanted to group by sex and find the number of rows of data for each sex, we would do:
surveys %>%
group_by(sex) %>%
tally
Here, tally()
is the action applied to the groups created by group_by()
and counts the total number of records for each category.
Challenge
How many individuals were caught in each
plot_type
surveyed?Use
group_by()
andsummarize()
to find the mean, min, and max hindfoot length for each species (usingspecies_id
).What was the heaviest animal measured in each year? Return the columns
year
,genus
,species_id
, andweight
.You saw above how to count the number of individuals of each
sex
using a combination ofgroup_by()
andtally()
. How could you get the same result usinggroup_by()
andsummarize()
? Hint: see?n
.
dplyr
is one part of a larger tidyverse
that enables you to work with data in tidy data formats. tidyr
enables a wide range of manipulations of the structure data itself. For example, the survey data presented here is in almost in what we call a long format - every observation of every individual is its own row. This is an ideal format for data with a rich set of information per observation. It makes it difficult, however, to look at the relationships between measurements across plots. For example, what is the relationship between mean weights of different genera across the entire data set?
To answer that question, we’d want each plot to have a single row, with all of the measurements in a single plot having their own column. This is called a wide data format. For the surveys
data as we have it right now, this is going to be one heck of a wide data frame! However, if we were to summarize data within plots and species, we might begin to have some relationships we’d want to examine.
Let’s see this in action. First, using dplyr
, let’s create a data frame with the mean body weight of each genera by plot.
surveys_gw <- surveys %>%
filter(!is.na(weight)) %>%
group_by(genus, plot_id) %>%
summarize(mean_weight = mean(weight))
head(surveys_gw)
spread
Now, to make this long data wide, we use spread
from tidyr
to spread out the different taxa into columns. spread
takes three arguments - the data, the key column, or column with identifying information, the values column - the one with the numbers. We’ll use a pipe so we can ignore the data argument.
surveys_gw_wide <- surveys_gw %>%
spread(genus, mean_weight)
head(surveys_gw_wide)
Notice that some genera have NA
values. That’s because some of those genera don’t have any record in that plot. Sometimes it is fine to leave those as NA
. Sometimes we want to fill them as zeros, in which case we would add the argument fill=0
.
surveys_gw %>%
spread(genus, mean_weight, fill = 0) %>%
head
We can now do things like plot the weight of Baiomys against Chaetodipus or examine their correlation.
surveys_gw %>%
spread(genus, mean_weight, fill = 0) %>%
cor(use = "pairwise.complete")
gather
What if we had the opposite problem, and wanted to go from a wide to long format? For that, we use gather
to sweep up a set of columns into one key-value pair. We give it the arguments of a new key and value column name, and then we specify which columns we either want or do not want gathered up. So, to go backwards from surveys_gw_wide
, and exclude plot_id
from the gathering, we would do the following:
surveys_gw_long <- surveys_gw_wide %>%
gather(genus, mean_weight, -plot_id)
head(surveys_gw_long)
Note that now the NA
genera are included in the long format. Going from wide to long to wide can be a useful way to balance out a dataset so every replicate has the same composition.
We could also have used a specification for what columns to include. This can be useful if you have a large number of identifying columns, and it’s easier to specify what to gather than what to leave alone. And if the columns are in a row, we don’t even need to list them all out - just use the :
operator!
surveys_gw_wide %>%
gather(genus, mean_weight, Baiomys:Spermophilus) %>%
head
Challenge
Make a wide data frame with
year
as columns,plot_id
as rows, and the values are the number of genera per plot. You will need to summarize before reshaping, and use the functionn_distinct
to get the number of unique types of a genera. It’s a powerful function! See?n_distinct
for more.Now take that data frame, and make it long again, so each row is a unique
plot_id
year
combination.The
surveys
data set is note truly wide or long because both there are two columns of measurement -hindfoot_length
andweight
. This makes it difficult to do things like look at the relationship between mean values of each measurement per year in different plot types. Let’s walk through a common solution for this type of problem. First, usegather
to create a truly long dataset where we have a key column calledmeasurement
and avalue
column that takes on the value of eitherhindfoot_length
orweight
. Hint: You’ll need to specify which columns are being gathered.With this new truly long data set, calculate the average of each
measurement
in eachyear
> for each differentplot_type
. Thenspread
them into a wide data set with a column forhindfoot_length
andweight
. Hint: Remember, you only need to specify the key and value columns forspread
.
Now that you have learned how to use dplyr
to extract information from or summarize your raw data, you may want to export these new datasets to share them with your collaborators or for archival.
Similar to the read.csv()
function used for reading CSV files into R, there is a write.csv()
function that generates CSV files from data frames.
Before using write.csv()
, we are going to create a new folder, data_output
, in our working directory that will store this generated dataset. We don’t want to write generated datasets in the same directory as our raw data. It’s good practice to keep them separate. The data
folder should only contain the raw, unaltered data, and should be left alone to make sure we don’t delete or modify it. In contrast, our script will generate the contents of the data_output
directory, so even if the files it contains are deleted, we can always re-generate them.
In preparation for our next lesson on plotting, we are going to prepare a cleaned up version of the dataset that doesn’t include any missing data.
Let’s start by removing observations for which the species_id
is missing. In this dataset, the missing species are represented by an empty string and not an NA
. Let’s also remove observations for which weight
and the hindfoot_length
are missing. This dataset should also only contain observations of animals for which the sex has been determined:
surveys_complete <- surveys %>%
filter(species_id != "", # remove missing species_id
!is.na(weight), # remove missing weight
!is.na(hindfoot_length), # remove missing hindfoot_length
sex != "") # remove missing sex
Because we are interested in plotting how species abundances have changed through time, we are also going to remove observations for rare species (i.e., that have been observed less than 50 times). We will do this in two steps: first we are going to create a dataset that counts how often each species has been observed, and filter out the rare species; then, we will extract only the observations for these more common species:
## Extract the most common species_id
species_counts <- surveys_complete %>%
group_by(species_id) %>%
tally %>%
filter(n >= 50)
## Only keep the most common species
surveys_complete <- surveys_complete %>%
filter(species_id %in% species_counts$species_id)
To make sure that everyone has the same dataset, check that surveys_complete
has 30463 rows and 13 columns by typing dim(surveys_complete)
.
Now that our dataset is ready, we can save it as a CSV file in our data_output
folder. By default, write.csv()
includes a column with row names (in our case the names are just the row numbers), so we need to add row.names = FALSE
so they are not included:
write.csv(surveys_complete, file = "data_output/surveys_complete.csv",
row.names = FALSE)
Data Carpentry,
2017. License. Contributing.
Questions? Feedback?
Please file
an issue on GitHub.
On
Twitter: @datacarpentry