Programming: functions and loops

Read the patients data into R environment
Writing our own functions
Doing repetitive tasks with loops

Read the patients data into R environment

patients <- read_tsv("data/patient-data-cleaned.txt") %>%  
    select(Height, Weight, -BMI)

Writing our own functions

Basic syntax to generate a function

Lets look at a tidyverse function definition

top_n

The command to create a function is function
This is followed by the arguments of the function in brackets
Finally the code that the function will execute is between curly brackets

So the general syntax is:

Function_Name <- function(arg1, arg2, arg3, ...){
    Code_Line_1
    Code_Line_2
    ...
}

Let’s create a simple function. This one has no arguments.

my_first_function  <- function(){
    message("Hello World!")
}

my_first_function()

## Hello World!

Let’s make it a bit more useful by adding an argument

sayHello <- function(name){
    message("Hello ", name, "!")
}

sayHello("Ash")

## Hello Ash!

sayHello("Matt")

## Hello Matt!

Functions with output

So far our function just printed a message to the console, but most functions return a value.

In R a function can only return a single object

Let’s create a function to calculate BMI: weight (kg) / height² (m) –> kg/m²

bmi <- function(height, weight){
    weight / (height ^ 2)
}
bmi(1.75, 78)

## [1] 25.46939

Let’s try that with our patients table.

patients  %>%  
    mutate(BMI=bmi(Height, Weight))

## # A tibble: 100 x 3
##   Height Weight     BMI
##    <dbl>  <dbl>   <dbl>
## 1   183.   76.6 0.00229
## 2   179.   80.4 0.00251
## 3   169.   75.5 0.00264
## # … with 97 more rows

This isn’t right because the height should be in metres. Let’s add some additional arguments to define the units and then some extra code to adjust the values depending on the units provided.

Defaults

bmi <- function(height, weight, height_units="m", weight_units="kg"){
    if(height_units == "cm"){ height <- height / 100 }
    if(weight_units == "g"){ weight <- weight / 1000 }
    weight / (height ^ 2)
}

bmi(1.75, 78)

## [1] 25.46939

patients  %>%  
    mutate(BMI=bmi(Height, Weight, height_units = "cm"))

## # A tibble: 100 x 3
##   Height Weight   BMI
##    <dbl>  <dbl> <dbl>
## 1   183.   76.6  22.9
## 2   179.   80.4  25.1
## 3   169.   75.5  26.4
## # … with 97 more rows

Returned object

There is a function return but we don’t usually need it.

bmi <- function(height, weight, height_units="m", weight_units="kg"){
    if(height_units == "cm"){ height <- height / 100 }
    if(weight_units == "g"){ weight <- weight / 1000 }
    bmi <- weight / (height ^ 2)
    return(bmi)
}
bmi(165, 67)

## [1] 0.002460973

Generating a plot with a function

diabetes <- read_tsv("data/diabetes.txt")

diabetes %>%  
    filter(ID == "AC/AH/001") %>%  
    ggplot(aes(x = Date, y = Glucose)) +
        geom_line(colour = "blue") +
        labs(title = "Patient: AC/AH/001")

How can we turn this into a function? Step by step…

patientID <- "AC/AH/001"
plotTitle <- str_c("Patient: ", patientID)

diabetes %>%  
    filter(ID == patientID) %>%  
    ggplot(aes(x = Date, y = Glucose)) +
        geom_line(colour = "blue") +
        labs(title = plotTitle)

plotGlucose <- function(patientID, dat=diabetes, myColour="blue"){
    plotTitle <- str_c("Patient: ", patientID)
    dat %>%  
        filter(ID == patientID) %>%  
        ggplot(aes(x = Date, y = Glucose)) +
            geom_line(colour = myColour) +
            labs(title = plotTitle)
}

plotGlucose("AC/AH/104")

plot1 <- plotGlucose("AC/AH/017", myColour="purple")
plot1

Doing repetitive tasks with loops

R works nicely with vectors

x <- c(1, 2, 3, 4, 5)
log2(x)

## [1] 0.000000 1.000000 1.584963 2.000000 2.321928

but not always the behaviour we want

sum(x, 5)

## [1] 20

so we could do…

sum(1, 5)
sum(2, 5)
sum(3, 5)
sum(4, 5)
sum(5, 5)

If you are repeating code like this, you should use loop.

Basic syntax

The basic syntax of a for loop is:

for(a_variable in a_vector){
    some code operating on a_variable
}

Let’s look at a simple example:

for(a_number in c(11, 21, 16, 4, 500)){
    sum(a_number, 5)
}

In this case the code has been run 5 times…

More commonly we would define the vector first:

some_data <- c(11, 21, 16, 4, 500)
for(a_number in some_data){
    sum(a_number, 5)
}

often we will use a sequence

some_data <- c(11, 21, 16, 4, 500)
for(a_number in seq(1, 5)){
    new_number <- sum(some_data[a_number], 5)
    message("Result number ", a_number, " is ", new_number)
}

## Result number 1 is 16

## Result number 2 is 26

## Result number 3 is 21

## Result number 4 is 9

## Result number 5 is 505

A better way is to use seq_along

some_data <- c(11, 21, 16, 4, 500)
for(a_number in seq_along(some_data)){
    new_number <- sum(some_data[a_number], 5)
    message("Result number ", a_number, " is ", new_number)
}

## Result number 1 is 16

## Result number 2 is 26

## Result number 3 is 21

## Result number 4 is 9

## Result number 5 is 505

A more realistic example

Let’s say that with our patients data we want to export a separate table for each state, containing just the men, with ID, Name, Smokes, BMI and their their average BP from the diabetes table.

states <- unique(patients$State)

avg_bp  <- diabetes %>%  
    group_by(ID) %>%  
    summarise(AvgBP=mean(BP)) %>%  
    ungroup()

for(state in states){
    outputFile <- str_c("data/", state, "_male_patient_data.tsv") %>%  
        str_replace(" ", "_")
    patients  %>%  
        filter(State == state & Sex == "Male") %>%  
        left_join(avg_bp, by = "ID") %>%  
        select(ID, Name, Smokes, BMI, AvgBP) %>%  
        write_tsv(outputFile)
}

Emma’s problem - A loop with plotting

There are many ways to solve this problem; this is just one of them.

all_patients <- unique(diabetes$ID)
len <- length(all_patients)
start_index <- seq(1, len, by = 16)

plotGlucose <- function(dat){
    ggplot(dat, aes(x = Date, y = Glucose)) +
        geom_line(colour = "red") +
        geom_point(shape = 21, fill = "red") +
        facet_wrap(~ID, ncol=4)
}

pdf("data/My_plots.pdf")
for(stt in start_index){
    end <- min(stt + 15, 100)
    these_patients <- all_patients[seq(stt, end)]
    diabetes %>%  
        filter(ID %in% these_patients) %>%  
        plotGlucose() %>%  
        print()
}
dev.off()

The last page isn’t ideal perhaps. We could solve this with some more advanced plotting using gridExtra to create the grid of plots from a list of individual plots. We will also need a nested loop - a loop within a loop.

library(gridExtra)

pdf("data/My_plots.pdf")
for(stt in start_index){
    end <- min(stt + 15, 100)
    these_patients <- all_patients[seq(stt, end)]
    plotList <- list() 
    for(i in seq_along(these_patients)){
        glPlot <- diabetes %>%  
            filter(ID %in% these_patients[i]) %>%  
            plotGlucose()
        plotList[[i]] <- glPlot
    }
    grid.arrange(grobs = plotList, ncol = 4, nrow = 4)
}
dev.off()

Programming: functions and loops – minimalscript

Ashley Sawle

Last modified: 07 Jun 2019