The following are instructions to download and manipulate the metadata from the "Mesophotic.org" repository in R (https://www.r-project.org/). Electronic notebook by Veronica Radice with contributions from Pim Bongaerts.
For these examples, please install (e.g. using install.packages(dplyr)) and load (as detailed below using require) the following packages for manipulating and viewing the data:
require(dplyr)
require(tidyr)
require(ggplot2)
require(ggstance)
require(ggpubr)
require(tidyverse)
require(cowplot)
You can import the latest database directly from the website into R:
publs <- read.csv("http://www.mesophotic.org/publications.csv",
as.is=TRUE)
# examine publications up until and including the year 2018 (eliminate the
# latest 2019 publications)
publs <- publs[!(publs$publication_year=="2019"),]
Show table dimensions and column names:
dim(publs)
colnames(publs)
Subset the publication list to those that are validated (i.e. metadata validated by two different content editors) or those that are included_in_stats (i.e. as validated but only including publications representing (peer-reviewed) scientific articles presenting original data from mesophotic depths):
publs_validated <- subset(publs, validated == "true")
publs_for_stats <- subset(publs, included_in_stats == "true")
Calculate the fold-increase in publications over the past 10 years (between 2008 and 2018):
publs_by_year <- publs_validated %>%
group_by(publication_year) %>%
summarise(count=n()) %>%
mutate(total_count = cumsum(count))
fold_increase <- publs_by_year[publs_by_year$publication_year==2018, "total_count"] /
publs_by_year[publs_by_year$publication_year==2008, "total_count"]
fold_increase
Where fields have multiple values (through so-called “HABTM” associations), such as research_fields , they are separated by semicolons, for example: Ecology; Community structure; Disturbances. Therefore, to query any of these fields, we need to separate these into individual rows. Then, we can plot for example the number of publications for each of those fields:
# Split `fields` into new rows
publs_fields <- separate_rows(publs_for_stats, "fields", sep=";\\s+")
# Remove rows with no associated `fields`
publs_field <- publs_fields[!(is.na(publs_fields$fields) |
publs_fields$fields == ""), ]
# Group by `fields`, count the number of publications and plot
top10_fields <- publs_fields %>%
group_by(fields) %>%
summarise(count=n()) %>%
top_n(10)
top10_fields$fields <- as.factor(top10_fields$fields)
top10_fields <- top10_fields %>% arrange(desc(count))
top10_fields
bar_fields <- ggplot(data = top10_fields, aes(x = count,
y = reorder(fields, count))) +
geom_barh(stat="identity") +
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
labs(x = "Publications", y = "Fields") + theme_bw()
bar_fields
# Split `focusgroups` into new rows
publs_focusgroups <- separate_rows(publs_for_stats, "focusgroups", sep=";\\s+")
# Remove rows with no associated `focusgroups`
publs_focusgroups <- publs_focusgroups[!(is.na(publs_focusgroups$focusgroups) |
publs_focusgroups$focusgroups ==""), ]
# count total number of records (rows)
countTotal_focusgroups <- nrow(publs_focusgroups)
count_focusgroups <- publs_focusgroups %>%
group_by(focusgroups) %>%
summarise(count=n())
# percent of focusgroup counts divided by total number of records of focusgroups (aka rows)
percent_focusgroups <- count_focusgroups %>%
mutate(percent=paste0(round(100*count/countTotal_focusgroups,1)))
percent_focusgroups$percent <- as.numeric(percent_focusgroups$percent)
top10_focusgroups <- percent_focusgroups %>%
top_n(10, percent) %>%
arrange(desc(count))
top10_focusgroups$focusgroups <- as.factor(top10_focusgroups$focusgroups)
top10_focusgroups
bar_focusgroups <- ggplot(data = top10_focusgroups, aes(x = percent,
y = reorder(focusgroups, count))) +
geom_barh(stat="identity") +
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
labs(x = "Percent", y = "Focus Groups") +
theme_bw()
bar_focusgroups
# Split `platforms` into new rows
publs_platforms <- separate_rows(publs_for_stats, "platforms", sep=";\\s+")
# Remove rows with no associated `platforms`
publs_platforms <- publs_platforms[!(is.na(publs_platforms$platforms) |
publs_platforms$platforms == ""), ]
# Group by `platforms`, count the number of publications and plot
top10_platforms <- publs_platforms %>%
group_by(platforms) %>%
summarise(count=n()) %>%
top_n(10)
top10_platforms$platforms <- as.factor(top10_platforms$platforms)
top10_platforms <- top10_platforms %>%
arrange(desc(count))
bar_platforms <- ggplot(data = top10_platforms, aes(x = count,
y = reorder(platforms, count))) +
geom_barh(stat="identity") +
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
labs(x = "Publications", y = "Platforms") +
theme_bw()
bar_platforms
# Split `locations` into new rows
publs_locations <- separate_rows(publs_for_stats, "locations", sep=";\\s+")
# Remove rows with no associated `locations`
publs_locations <- publs_locations[!(is.na(publs_locations$locations) |
publs_locations$locations == ""), ]
# Group by `locations`, count the number of publications and plot
top10_locations <- publs_locations %>%
group_by(locations) %>%
summarise(count=n()) %>%
top_n(10)
top10_locations$locations <- as.factor(top10_locations$locations)
top10_locations <- top10_locations %>% arrange(desc(count))
bar_locations <- ggplot(data = top10_locations, aes(x = count,
y = reorder(locations, count))) +
geom_barh(stat="identity") +
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
labs(x = "Publications", y = "Locations") +
theme_bw()
bar_locations
#pdf("Fig category bargraphs.pdf", width = 9, height = 6,
encoding="MacRoman")
plot <- plot_grid(bar_focusgroups,
bar_fields,
bar_platforms,
bar_locations,
labels = c('a)', 'b)', 'c)', 'd)'),
hjust = -1,
vjust = 1,
align = 'vh',
scale = c(1, 1, 1, 1),
label_size = 12,
ncol = 2,
label_fontface = "plain")
plot
#dev.off()
Data for the visualizations on the http://mesophotic.org/stats page can be downloaded as CSVs and manipulated too. For example:
pubs_over_depth <- read.csv("pubs_over_depth.csv", as.is = TRUE)
pubs_over_depth <- data.frame(pubs_over_depth$Publications, pubs_over_depth$Category)
Publications by max depth
colnames(pubs_over_depth) <- c("Publications", "Depth")
#pdf("Fig Publications over depth.pdf", width = 8, height = 6, encoding="MacRoman")
g <- ggplot(pubs_over_depth, aes(x = Depth, y = Publications)) +
geom_area(color="#69b3a2", fill = "#69b3a2") +
scale_x_reverse(breaks = seq(150, 30, -8), limits = c(150, 30)) +
scale_y_continuous(breaks = seq(0, 500, 50), limits = c(0, 500)) +
labs(x = "Depth (m)", y = "Publications") +
theme_classic()
g + rotate()
#dev.off()
Import data Publications over time
pubs_over_time <- read.csv("pubs_over_time.csv", as.is = TRUE)
colnames(pubs_over_time) <- c("year", "value")
# modify data frame by creating new variable
pubs_over_time <- mutate(pubs_over_time, category = "Publication")
pubs_over_time$category <- as.factor(pubs_over_time$category)
pubs_over_time
Import data Locations (by Year) over time
locations_over_time <- read.csv("locations_over_time.csv", as.is = TRUE)
colnames(locations_over_time) <- c("year", "value")
# modify data frame by creating new variable
locations_over_time <- mutate(locations_over_time,
category = "Study location")
locations_over_time$category <- as.factor(locations_over_time$category)
locations_over_time
Import data First Authors (by Year) over time
people_over_time <- read.csv("people_over_time.csv", as.is = TRUE) # unique first authors
colnames(people_over_time) <- c("year", "value")
# modify data frame by creating new variable
people_over_time <- mutate(people_over_time, category = "First author")
people_over_time$category <- as.factor(people_over_time$category)
people_over_time
Import data term "mesophotic" (by Year) over time
mesophotic_over_time <- read.csv("mesophotic-mentions-cumulative.csv",
as.is = TRUE)
mesophotic_over_time <- mesophotic_over_time[-c(3)]
colnames(mesophotic_over_time) <- c("year", "value")
# modify data frame by creating new variable
mesophotic_over_time <- mutate(mesophotic_over_time,
category = "Term 'mesophotic'")
mesophotic_over_time$category <- as.factor(mesophotic_over_time$category)
mesophotic_over_time
Merge data that look at counts (Pubs, Location, etc) over time (by Year)
# join data
# append to dataframe as new rows
data_over_time <- bind_rows(locations_over_time, pubs_over_time,
mesophotic_over_time, .id = NULL) # could add people_over_time
data_over_time$category <- as.factor(data_over_time$category)
data_over_time
Publications and Locations over time Stacked Area chart
# designate specific order
data_over_time$category <- factor(data_over_time$category,
levels = c("Publication", "Study location", "Term 'mesophotic'")) # could add "First author"
# colorblind friendly palette
# http://www.cookbook-r.com/Graphs/Colors_(ggplot2)/
# cbPalette <- c("#999999", "#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2", "#D55E00", "#CC79A7")
cbPalette <- c("#0072B2", "#56B4E9", "#999999", "#E69F00") # "#E69F00", "#009E73",
# alternate color pallets via color brewer
# scale_fill_brewer(palette="Greens")
#pdf("Fig Series of data over time.pdf", width = 8, height = 5, encoding="MacRoman")
ggplot(data_over_time, aes(x=year, y=value, fill=category)) +
geom_area(colour="black", size=.3, alpha=.4) +
scale_fill_manual(values = cbPalette) +
scale_x_continuous(name = "Year", breaks = seq(1990, 2018, 4), limits = c(1990, 2018)) +
ylab("Number") +
labs(fill = "Series") +
theme_classic() +
theme(legend.title = element_text(), legend.justification = c(0, 1), legend.position = c(0, 1), legend.box.margin=margin(c(5,5,5,5)))
#dev.off()