Bar chart of common mental disorders...

My day job in the School of Medicine at Cardiff University involves facilitating learning around various medical conditions including mental health. I like a few statistics so I have been exploring the prevalence of mental health disorders. I found a report about mental health from Our World in Data which shares all the data it uses on Github - making it open source. There is lots of interesting data.
As well as mental health, there is data and reports about cancer and the burden of disease.

Inspired by the mental health report from Our World in Data, I downloaded some data and generated a graph which shows the prevalence of Mental Health Disorders in the UK.

Here is the graph:

Here is the R script that generated the graph and a few other graph along the way.

===  START ===
# looking at some mental health data...
# source: https://ourworldindata.org/mental-health

library(readr)
library(dplyr)
library(tidyr)
library(ggplot2)

# download the data from Github
data <- read_csv("https://raw.githubusercontent.com/owid/owid-datasets/master/datasets/Mental%20health%20prevalence%20(IHME)/Mental%20health%20prevalence%20(IHME).csv")

# pull out data for UK and wrangle using pipes and dplyr
data %>% 
    # filter() by country and year
    filter(Entity == "United Kingdom", Year == 2016) %>%
    # select() prevalence - percentage 3rd to 13th column
    select(3:13) %>%
    # turn from wide format to long for better plotting using gather()
    gather(key = "CMHD", value = "prevalence") -> data1

# now have new object data1

# first bar chart...
ggplot(data1, aes(x = CMHD, y = prevalence)) +
    geom_bar(stat = "identity")

# plot horizontally with coord_flip()
ggplot(data1, aes(x = CMHD, y = prevalence)) +
    geom_bar(stat = "identity") +
    coord_flip()

# remove the text "- both sexes (percent)" gsub() function
data1$CMHD <- gsub(" \\- both sexes \\(percent\\)", "", data1$CMHD)
# the \\ are escape characters for minus and brackets 

# AND

# reorder the categories as factors by size of prevalence
# https://www.reed.edu/data-at-reed/resources/R/reordering_geom_bar.html
data1$CMHD <- factor(data1$CMHD, levels = data1$CMHD[order(data1$prevalence)])

p <- ggplot(data1, aes(x = CMHD, y = prevalence)) +
    geom_bar(stat = "identity") +
    coord_flip()
p

# add some labels and source....
p <- p +
    theme_bw() +
    labs(x = "",
        y = "Prevalence (%)",
        title = "Prevalence of Common Mental Health Disorders in UK (2016)", 
        subtitle = "https://ourworldindata.org/mental-health")
p

# Our World in Data website has the numbers on the plot...
p <- p +
    geom_text(aes(label=round(prevalence, 2)))
p

# Our World website has different coloured bars on the plot...
# by altering fill in the aes() of ggplot
p <- ggplot(data1, aes(x = CMHD, y = prevalence, fill = CMHD)) +
    geom_bar(stat = "identity") +
    coord_flip() +
    theme_bw() +
    labs(x = "",
        y = "Prevalence (%)",
        title = "UK Prevalence of Common Mental Health Disorders (2016)", 
        subtitle = "https://ourworldindata.org/mental-health") +
    geom_text(aes(label=round(prevalence, 1)))
p

#  Which adds a legend... so remove the legend...
p + theme(legend.position="none")

=== END ===

Some resources:

• Exploring diseases in Wales for SQL Relay...
• A bar chart for Athena SWAN awards
• ggplot documentation for geom_bar()
• http://www.sthda.com/english/wiki/ggplot2-barplots-quick-start-guide-r-software-and-data-visualization
• Another source of data: https://www.mentalhealth.org.uk/statistics

Exploring the UK Gender Pay Gap with R...

The gender pay gap in the UK may not of primary interest to some biochemists but the Biochemical Society is interested in gender equality and the majority of biochemistry undergraduates are female... Here is the Biochemical Society's policy statement and here is something relevant from their blog.

A report about the gender pay gap was tweeted about today by @UKParliament (it seems it was published last year). There is an Excel file that goes with the report.

Today, I've been using R to explore some of the data and I have written a script below to make these graphs - the first two graphs from the report.

The way the data is presented makes me uncomfortable with men being paid more represented as a positive percentage and women being paid more being expressed as a negative percentage. I feel sure there is a better way....

Still, the data is interesting....

Here is the script:
START
library(RCurl)
library(readxl)
library(ggplot2)
library(reshape2)
library(ggthemes)

# this is the link to the data
link <- "http://researchbriefings.files.parliament.uk/documents/SN07068/data-tables.xlsx"

# the download.file() function downloads and saves the file with the name given
download.file(url=link,destfile="file.xlsx", mode="wb")

# then we can open the file and extract the data using the read_excel() function. 
data <- read_excel("file.xlsx", skip=3, col_names=TRUE)

str(data)
# shows that Year is characters 
data[,1] <- as.numeric(data[,1])   # change to number
data <- data[1:22,]    # get rid of seven rows of NAs.
names <- colnames(data)
names[1] <- "Year"
names[2] <- "All_employees"
colnames(data) <- names  # make column names easier to use
data[,2:4] <- data[,2:4]*100   # Excel stores percents as decimals

# reshape the data from wide to long format
data.melt <- melt(data, id.vars = "Year")
colnames(data.melt) <- c("Year", "empType", "gendGap")


# draw the graph
p1 <- ggplot(data.melt, aes(x=Year, 
                           y= gendGap, 
                           colour = empType)) + 
  geom_point() +   # draw the points
  geom_line(size=1) +  # draw the lines
  labs(color = "Employment Type") + # customizes the legend title
  ylab("Gender Gap (%)") + # y-label
  ggtitle("Gender Pay Gap, UK, 1997-2015") +   # graph title
  ylim(-10,30) + 
  xlim(1995, 2015) +
  geom_hline(yintercept = 0) +  # nice line at zero
  theme_bw()

p1 <- p1 + theme(legend.text=element_text(size = 12), # increase size of text
               legend.title=element_text(size = 12)) # and title

p1 <- p1 + theme(axis.title.y = element_text(size = 14 )) + 
  theme(axis.text = element_text(size = 12))

p1 # show the graph

# maybe you prefer a different theme.
p1 + theme_hc()

# maybe without a legend but with labels on the lines:
p1 <- p1 + theme(legend.position="none") + 
  geom_text(data = data.melt[which(data.melt$Year == "2013"),],
               aes(label = empType),
               vjust = -2)
p1

# draw the second graph with the age data... 
data2 <- read_excel("file.xlsx", sheet=2, skip=3, col_names=TRUE)
View(data2)
str(data2)
data2 <- data2[1:8,]

# multiply numbers by 100 to give percentages
data2[,2:4] <- data2[,2:4]*100

names <- colnames(data2)
names[1] <- "Ages"
names[2] <- "All_employees"
colnames(data2) <- names
data2.type <- data2[3:8,]

data2.melt <- melt(data2.type, id.vars = "Ages")
colnames(data2.melt) <- c("Ages", "empType", "gendGap")

g <- ggplot(data = data2.melt[7:18,], aes (x = Ages, y = gendGap, fill = empType))  
g <- g + geom_bar(stat="identity", position="dodge", width = 0.75) +
  ylim(-11,20)  +
  ylab("Gender Gap (%)") + # y-label
  xlab("Age") + # x-label
  ggtitle("Gender Pay Gap by Age, April 2015") +
  labs(fill = "Employment Type") + # customizes the legend title
  theme_hc() +
  theme(legend.position=c(0,1), # move to the top left
        legend.justification=c(0,1.5)) # move it in a bit
g  # show the graph...

Overview of gender breakdown across types of staff

I have spent a large part of the last month grappling with Athena SWAN data for the School of Medicine at Cardiff University. To apply for this equality award, the Athena SWAN Self Assessment Team needed to look at the gender breakdown of students and staff in the School.

Here is a bar chart I generated with the data to give an overview of the School:

Here's is the script I used to generate this data and some other graphs that explore using ggplot to make bar charts.

START
# overview of the School in terms of gender for our Athena SWAN application
library(readxl)
library(ggplot2)
library(reshape2)

# the data is here
link <- "https://raw.githubusercontent.com/brennanpincardiff/AthenaSWANBenchmarkData/master/asStaffOverview_comparedtype_20160505.xlsx"

# the download.file() function downloads and saves the file with the name given
download.file(url=link, destfile="file.xlsx", mode="wb")

# read in the data...
data <- read_excel("file.xlsx")

# calculate Female and Male as a percentage
data$tot <- data$f + data$m
data$Female <- (data$f / data$tot) *100 
data$Male <- (data$m / data$tot) *100

# first a simple plot of the Cardiff School of Medicine Data
# plot the number of females of each type of member of the School
# pull out the Cardiff data
data.c <- data[data$uni == "Cardiff",2:5]

# make the object with the Cardiff subset  
g <- ggplot(data.c,
       aes(x = type,
           y = f)) +
     geom_bar(stat="identity")+
     theme(axis.text.x = element_text(size = 12))

# show the graph
g

# The order of the bars is not what I want.
# take the order from the data file
types <- data.c$type

# apply this order to the graph
# http://stackoverflow.com/questions/5208679/order-bars-in-ggplot2-bar-graph
g + scale_x_discrete(limits = types)

# I want to put female and male on the same graph so...
# pull out and melt the Cardiff data so that have f and m for each type in long format
data.melt <- melt(data[data$uni == "Cardiff",2:5], 
                  id.var = c("type", "year"))

# put the data in an object
g <- ggplot(data.melt, 
       aes(x = type, 
           y = value,
           fill = variable)) + 
      scale_x_discrete(limits = types)

# show with geom_bar
# this is a stacked bar chart showing both female and male numbers
g + geom_bar(stat="identity")

# show female and male in separate bars using position arguement
g + geom_bar(stat="identity", position="dodge")

# have as stacked bar chart but normalised to one. 
g +  geom_bar(stat="identity", position="fill")

# I really want a percentage of each stacked 
# to allow different parts of the School to be compared
# remove the raw numbers
data.s <- data[, -c(3:6)]
data.melt.s <- melt(data.s[data.s$uni == "Cardiff",2:4], 
                  id.var = c("type"))  # id = or id.var are same
colnames(data.melt.s) <- c("type", "Gender", "Percent")

# make a new graph
p <- ggplot(data.melt.s, 
           aes(x = type, 
               y = Percent,
               fill = Gender)) +
      geom_bar(stat="identity") + 
      scale_x_discrete(limits = types) +
      ylab("Percentage of staff") +
      xlab("") +
      theme_bw() +
      theme(axis.text.x = element_text(size = 14))
p # show the graph

# adjust the colours using scale_fill_brewer()
p + scale_fill_brewer(palette = 16, direction=-1) +
  # and add a title
    ggtitle("Gender breakdown of different members of the School of Medicine") +
    theme(axis.title.y = element_text(size = 14 ))

# use data from Imperial & Leeds as comparison
# use the facet_wrap function to make all looks the same. 
# remove Prof & Support as we only have for Cardiff
data.s <- data.s[-5,]
types <- types[-5]

# melt the data into the format for the bar chart. 
data.melt.s <- melt(data.s, 
                    id.var = c("uni", "type"))
colnames(data.melt.s) <- c("uni", "type", "Gender", "Percent")

# this is a final plot with three Medical Schools (it's a bit big). 
p <- ggplot(data.melt.s, 
            aes(x = type, 
                y = Percent,
                group = uni,
                fill = Gender)) +
  geom_bar(stat="identity") +
  facet_wrap(~uni) +    # this function gives three plots
  scale_x_discrete(limits = types) +
  ylab("Percentage of staff") +
  xlab("") +
  theme_bw() +
  scale_fill_brewer(palette = 16, direction=-1) +
  ggtitle("Gender breakdown of three Schools of Medicine") +
  theme(strip.text.x = element_text(size = 14, colour = "black")) +
  theme(axis.text.x = element_text(size = 11))

p # show the plot

END of SCRIPT

I looked up these pages for help:

• http://stackoverflow.com/questions/3290330/facet-label-font-size
• http://www.sthda.com/english/wiki/ggplot2-axis-ticks-a-guide-to-customize-tick-marks-and-labels

A bar chart looking at the number of Athena SWAN awards...

Bar charts are commonly used. They are easy to understand. They allow us to present data is a visual way. Lots of people appreciate data visually and prefer pictures to numbers. I don't just analyse biological data. So here is an example of using ggplot to create a stacked bar chart to show the increase in Athena SWAN Awards over the last six years.

Here is the graph:

Here is the script:

# activate the required packages
library(ggplot2)
library(reshape2)
library(ggthemes)

# here is the data - gathered from various awards booklets
# and the ECU press release
years <- c("2009","2010","2011","2012","2013", "2014")
bronze <- c(19,  13, 25, 66, 135, 152)
silver <- c(16, 16, 14, 26, 40, 43)
gold <- c(0,1,0,2,4,0)

# create the data frame required for ggplot
swan.df <- as.data.frame(years)
swan.df$Bronze <- bronze
swan.df$Silver <- silver
swan.df$Gold <- gold

# reshape the data from long into short format
as.melt <- melt(swan.df, id.vars = "years", value.name = "number", variable.name = "Awards")

# make the first version of the plot
p <- ggplot(as.melt, aes(x=years, y=number, fill=Awards)) + 
     geom_bar(stat="identity") +  # this bit makes the barplot
     scale_fill_manual(values=c("brown", "grey", "yellow")) + # control the colours
     xlab("") + #no need for the "years" x-axis
     theme_few() # nice clean theme  

# I want to add a nice y-axis label
# and increase the size of the text
p <- p + ylab("Number of Awards") +
     theme(axis.title.y = element_text(size = 14 )) + 
     theme(axis.text = element_text(size = 12))

# make the legend a bit bigger and move it to top left 
p <- p + theme(legend.position=c(0,1), # moves it to the top left
               legend.justification=c(0,1), # moves it in a bit
               legend.text=element_text(size = 12), # increase the size of the title
               legend.title=element_text(size = 12)) # and the labels

# have a look at the bar chart
p

# save the graph....
p + ggsave("AthenaSWANawards09_14.pdf")

END of SCRIPT

Important resource: http://www.cookbook-r.com/Graphs/Bar_and_line_graphs_(ggplot2)/

Going beyond bar charts.

Bar charts are not very data rich. There is often a better way to show the data or a better way to do the experiment. As a student and post-doctoral fellow, I was encouraged to look to investigate relationships in more detail. For example, it's better to do a time course experiment or to investigating dose relationships - or both! These suggest a more detailed investigation of a biological system. This would be better plotted in other ways. For examples, see the protein assay graph and the LD50 graph.