Programming with data in R
Susan Holmes (c)
Logical statements
Suppose I use the data on births from fivethirtyeight we
looked at in the last session. I would like to extract all
the children born on Friday the 13th.
Download the "births.csv" data from our folder:
https://stanford.box.com/s/jm9pe4c4ve1kuw2kenlxohogo56ogsve
You can place the data directly in your working directory to make it
easier to use. If you are not sure where your current working directory
is, you can always check your directory using command getwd().
load("births.RData")
head(births)
## year month date_of_month day_of_week births
## 1 2000 1 1 6 9083
## 2 2000 1 2 7 8006
## 3 2000 1 3 1 11363
## 4 2000 1 4 2 13032
## 5 2000 1 5 3 12558
## 6 2000 1 6 4 12466
summary(births)
## year month date_of_month day_of_week
## Min. :2000 Min. : 1.000 Min. : 1.00 Min. :1
## 1st Qu.:2003 1st Qu.: 4.000 1st Qu.: 8.00 1st Qu.:2
## Median :2007 Median : 7.000 Median :16.00 Median :4
## Mean :2007 Mean : 6.523 Mean :15.73 Mean :4
## 3rd Qu.:2011 3rd Qu.:10.000 3rd Qu.:23.00 3rd Qu.:6
## Max. :2014 Max. :12.000 Max. :31.00 Max. :7
## births
## Min. : 5728
## 1st Qu.: 8740
## Median :12343
## Mean :11350
## 3rd Qu.:13082
## Max. :16081
I can see that the data on day of the week is not a factor
but a numeric encoding, in fact Friday is encoded as 5.
Fridays=births[which(births[,4]==5),]
Which of the Friday births occurred on the 13th?
Fridays13=Fridays[which(Fridays[,3]==13),]
head(Fridays13)
## year month date_of_month day_of_week births
## 287 2000 10 13 5 11723
## 469 2001 4 13 5 10881
## 560 2001 7 13 5 12187
## 987 2002 9 13 5 13028
## 1078 2002 12 13 5 11600
## 1260 2003 6 13 5 12013
dim(Fridays13)
## [1] 25 5
Weekendbirths=births[which(births[,4]%in%c(6,7)),]
Weekdaybirths=births[which(births[,4]<6),]
Valid ways of generating TRUE or FALSE
Sign Meaning Example
== Equals day_of_week == 5
!= Does not equal year != 0
Sign Meaning Example
> Greater than day_of_week > 5
>= Greater than or equal date_of the week >= 6
< Less than day_of_week <2
<= Less than or equal to day_of_week <=1
%in% Included in births[,4]%in%c(6,7)
is.na() Is a missing value is.na(births[,4])
Conditions and directions
if (condition){
Do something
} else {
Do something different
}
if (mean(Weekendbirths[,5])> mean(Weekdaybirths[,5]))
{ cat("More weekend babies on average") }else {
cat("There are less weekend babies on average")
}
## There are less weekend babies on average
Loops and repeats
One may want to repeat a computation for each
different element of a vector, sometimes we need to do
this with loops and sometimes we avoid this with what is
called vectorization.
A typical loop:
set.seed(431)
mat43=replicate(4,sample(3,3))
mat43l=rep(0,4)
for (j in 1:4)
{mat43l[j]=max(mat43[,j])}
mat43l
## [1] 3 3 3 3
A vectorized version:
apply(mat43,2,max)
## [1] 3 3 3 3
Why is vectorization faster, since the number of
operations seems always the same?
apropos("apply")
## [1] ".mapply" "apply" "dendrapply" "eapply" "kernapply"
## [6] "lapply" "mapply" "rapply" "sapply" "tapply"
## [11] "vapply"
?apply
Because R is interpreted it deals with assigning types and
memory to variables on the fly. The inner representation
of every variable is a vector, it expects to act on vectors
even if only one number is involved.
Actual example with larger numbers:
mat43=replicate(5000,sample(30000,1000))
dim(mat43)
## [1] 1000 5000
system.time(apply(mat43,2,max))
## user system elapsed
## 0.061 0.009 0.070
mat43l=rep(0,5000)
system.time(for (j in 1:5000)
{mat43l[j]=max(mat43[,j])})
## user system elapsed
## 0.041 0.003 0.044
Functions in R
R is called a functional programming language because
the actions we take are done using functions, even
quitting at the end is done using q().
Example
library(readxl)
read_excel
## function (path, sheet = 1, col_names = TRUE, col_types = NULL,
## na = "", skip = 0)
## {
## path <- check_file(path)
## ext <- tolower(tools::file_ext(path))
## switch(excel_format(path), xls = read_xls(path, sheet, col_names,
## col_types, na, skip), xlsx = read_xlsx(path, sheet, col_names,
## col_types, na, skip))
## }
## <environment: namespace:readxl>
Writing our own functions
Useful if you are execute the same set of commands on
different data or with different parameters.
vec=c(1,2,3,5,7,11,13,17,19,23)
(vec^2)+ 1
## [1] 2 5 10 26 50 122 170 290 362 530
(vec^2)+ 3
## [1] 4 7 12 28 52 124 172 292 364 532
(vec^3)
## [1] 1 8 27 125 343 1331 2197 4913 6859 12167
Suppose we wanted to take another vector and see
which of its elements were divisible by 2,3,5, or other
numbers.
We write a function
ExpAnd <- function(vec,exponent,addto)
{
vec^exponent+addto
}
We will put it and edit it in the top part of our RStudio
IDE and save it as a file that we can also source later.
ExpAnd <- function(vec,exponent,addto)
{
out=vec^exponent+addto
return(out)
}
ExpAnd <- function(vec,exponent,addto)
{##Function that takes argument vec to the power
## exp, adds add and then outputs the result
out=vec^exponent+addto
return(out)
}
Question What happens if you apply your function with
vec = 3, exponent=4, and addto=4 ?
Question What happens if you type:
ExpAnd()
Put some default values in the function:
ExpAnd <- function(vec=seq(4,25,3),exponent=2,addto=3)
{
# Function that takes argument vec to the power
# exp, adds add and then outputs the result
out<-vec^exponent+addto
return(out)
}
ExpAnd()
## [1] 19 52 103 172 259 364 487 628
Question: Try calling the function ExpAnd with
ExpAnd(vec=seq(4,25,3),exponent=2,addto="3")
ExpAnd(vec=seq(4,25,3),exponent=2,addto="3")
ExpAnd <- function(vec=seq(4,25,3),exponent=2,addto=3){
# Function that takes argument vec to the power
# exp, adds add and then outputs the result
if (any(!is.numeric(c(vec,exponent,addto))))
stop("One of the arguments is not numeric.")
out <- vec^exponent+addto
return(out)
}
What happens if we now type:
ExpAnd(addto="4")
Functions are quite robust to some changes in the input:
ExpAnd(vec=matrix(c(2,3,4,1,1,2,2,7),ncol=2),3,0)
## [,1] [,2]
## [1,] 8 1
## [2,] 27 8
## [3,] 64 8
## [4,] 1 343
Note: Passing an unspecified number of parameters to a
function
We can pass extra, unspecified arguments to a function
by using the ... notation in the argument list.
add20 <- function(x, ...) {
k <- x+20
return(k)
}
Summary of this Session:
We have introduced the notion of logical
variables that test certain facts.
We saw how to combine the function which
with a logical statement to take a subset of the
data.
We can execute blocks of commands
encapsulated with {} using if and else with
logical conditions.
We saw that functions are an important
component of R programming.
Their basic elements are a name, an argument
and an output that is returned using the return
function.
We edit functions in an external file and can call
them in later session by using the function
source().
We can test the flow of the function and stop it if
something goes awry.
Question: Go to the cheatsheet: Base R
Look at all the functions we have not tried yet and try
the examples.
