It is important to understand that there are many different ways to achieve the same thing, particularly in base R. I have tried to use simpler syntax where possible. For the more complicated cases, I have based my solutions on top hits from web searches.
It is also worth remembering that when you look for help online, you might find solutions using “base R” syntax. If you are used to “tidyverse” packages, it will help to add keywords such as “dplyr” and “tidyverse” to your search.
There are many other resources available online, in particular several cheatsheets that might be useful as a reference: - https://www.rstudio.com/resources/cheatsheets/
Setup Load the packages needed for these examples:
library(dplyr)
library(tidyr)The datasets used in these examples come pre-loaded with R. To find more about them check their help pages:
# ?iris
# ?mtcars
# ?Indomethselection <- select(iris, Species, Petal.Width) # by name
head(selection)selection <- select(iris, 5, 4) # by column index
head(selection)selection <- iris[, c("Species", "Petal.Width")] # by name
head(selection)selection <- iris[, c(5, 4)] # by column index
head(selection)mutated <- mutate(iris,
Petal.Ratio = Petal.Length/Petal.Width,
Sepal.Ratio = Sepal.Length/Sepal.Width)
head(mutated)mutated <- iris$Petal.Ratio <- iris$Petal.Length/iris$Petal.Width
mutated <- iris$Sepal.Ratio <- iris$Sepal.Length/iris$Sepal.Width
head(mutated)## [1] 1.457143 1.633333 1.468750 1.483871 1.388889 1.384615filtered <- filter(iris, Petal.Width > 0.5 & Species == "setosa")
head(filtered)# Using [,]
filtered <- iris[iris$Petal.Width > 0.5 & iris$Species == "setosa", ]
head(filtered)# Using subset (works very much like dplyr::filter)
filtered <- subset(iris, Petal.Width > 0.5 & Species == "setosa")
head(filtered)# descending order of species (alphabetic) followed by ascending order of Petal.Width
arrange(iris, desc(Species), Petal.Width) # descending order of species (alphabetic) followed by ascending order of Petal.Width
iris[order(rev(iris$Species), iris$Petal.Width) , ]# Generic way
summarise(iris,
Petal.Length.mean = mean(Petal.Length),
Petal.Length.sd = sd(Petal.Length),
Sepal.Length.mean = mean(Sepal.Length),
Sepal.Length.sd = sd(Sepal.Length))# Shortcut when same functions applied to same variables
summarise_at(iris,
.vars = c("Petal.Length", "Sepal.Length"),
.funs = c("mean", "sd"))(see also variants summarise_all() and summarise_if())
There are many ways to do this with base R, here’s a couple of possibilities:
# Manually create a data.frame
data.frame(Petal.Length.mean = mean(iris$Petal.Length),
Petal.Length.sd = sd(iris$Petal.Length),
Sepal.Length.mean = mean(iris$Sepal.Length),
Sepal.Length.sd = sd(iris$Sepal.Length))# Use the "aggregate" function
## Column names might have to be changed afterwards
aggregate(formula = c(Sepal.Length, Petal.Length) ~ 1,
data = iris,
FUN = function(x) c(mean = mean(x), sd = sd(x)))Consider the following tables
band_members
## # A tibble: 3 x 2
## name band
## <chr> <chr>
## 1 Mick Stones
## 2 John Beatles
## 3 Paul Beatles
band_instruments
## # A tibble: 3 x 2
## name plays
## <chr> <chr>
## 1 John guitar
## 2 Paul bass
## 3 Keith guitar
band_instruments2
## # A tibble: 3 x 2
## artist plays
## <chr> <chr>
## 1 John guitar
## 2 Paul bass
## 3 Keith guitar# Retain rows with matches in both tables
inner_join(band_members, band_instruments, by = "name") # Retain all rows:
full_join(band_members, band_instruments, by = "name") # Retain all rows from first table:
left_join(band_members, band_instruments, by = "name") # Retain all rows from second table:
right_join(band_members, band_instruments, by = "name") Columns used for merging can have different names between tables:
inner_join(band_members, band_instruments2,
by = c("name" = "artist"))# Retain rows with matches in both tables
merge(band_members, band_instruments, by = "name") # Retain all rows:
merge(band_members, band_instruments, by = "name", all = TRUE) # Retain all rows from first table:
merge(band_members, band_instruments, by = "name", all.x = TRUE) # Retain all rows from second table:
merge(band_members, band_instruments, by = "name", all.y = TRUE) Columns used for merging can have different names between tables:
merge(band_members, band_instruments2, by.x = "name", by.y = "artist")Important note: with dplyr, grouped operations are initiated with the function group_by(). It is a good habit to use ungroup() at the end of a series of grouped operations, otherwise the groupings will be carried in downstream analysis, which is not always desirable. In the examples below we follow this convention.
Note on base R: with base R there are several ways to do these types of operations. Here we show a generic “split-apply-combine” solution using the by() function. These solutions require some understanding of list objects and how to write custom functions.
mtcars %>%
group_by(cyl, gear) %>%
summarise(mpg.mean = mean(mpg),
mpg.sd = sd(mpg),
wt.mean = mean(wt),
wt.sd = sd(wt)) %>%
ungroup() # remove any groupings from downstream analysis## `summarise()` has grouped output by 'cyl'. You can override using the `.groups` argument.# First operate in the data.frame by group (split-apply)
mtcars_by <- by(mtcars,
INDICES = list(mtcars$cyl, mtcars$gear),
FUN = function(x){
data.frame(cyl = unique(x$cyl),
gear = unique(x$gear),
mpg.mean = mean(x$mpg),
mpg.sd = sd(x$mpg),
wt.mean = mean(x$wt),
wt.sd = sd(x$wt))
})
# Then combine the results into a data.frame
do.call(rbind, mtcars_by)Alternative solution:
# Using aggregate
aggregate(formula = cbind(mpg, wt) ~ cyl + gear,
data = mtcars,
FUN = function(x){
c(mean = mean(x), sd = sd(x))
})For example, center the measurements of “Petal.Width” by subtracting the mean within species.
iris %>%
group_by(Species) %>%
mutate(Petal.Width.centered = Petal.Width - mean(Petal.Width)) %>%
ungroup() # remove any groupings from downstream analysis# First operate in the data.frame by group (split-apply)
iris_by <- by(iris,
INDICES = iris$Species,
FUN = function(x){
x$Petal.Width.centered <- x$Petal.Width - mean(x$Petal.Width)
return(x)
})
# Then combine the results into a data.frame
do.call(rbind, iris_by)Alternative solution:
iris$Petal.Width.centered <- ave(iris$Petal.Width, iris$Species, FUN = function(x) x - mean(x))Filter rows with conditions evaluated within groups iris flowers with maximum “Petal.Width” for each “Species”.
iris %>%
group_by(Species) %>%
filter(Petal.Width == max(Petal.Width))# First operate in the data.frame by group (split-apply)
widest_petals <- by(iris,
INDICES = iris$Species,
FUN = function(x){
x[x$Petal.Width == max(x$Petal.Width), ]
})
# Then combine the results into a data.frame
do.call(rbind, widest_petals)gather(iris, key = "trait", value = "measurement",
Sepal.Length:Petal.Width)I couldn’t find an easy way to do this in a clean way, although reshape() gets us close (but the “trait” variable is coded numerically):
reshape(iris,
varying = c("Sepal.Length", "Sepal.Width",
"Petal.Length", "Petal.Width"),
timevar = "trait",
idvar = "id",
v.names = "measurement",
direction = "long")There is also stack(), but we loose the “Species” column in the process.
stack(iris, select = c("Sepal.Length", "Sepal.Width",
"Petal.Length", "Petal.Width"))spread(Indometh, key = "time", value = "conc")Note that in this case it might have been better to first modify the “key” column so that the resulting column names are not numeric:
Indometh %>%
mutate(time = paste("conc", time, sep = "_")) %>%
spread(key = "time", value = "conc")reshape(Indometh,
v.names = "conc",
idvar = "Subject",
timevar = "time",
direction = "wide")Copyright © 2021 Carlos Sanchez.