Justin Varholick 2024-01-12
After you have analyzed your dominance data (e.g., cleaned, wrangled, and calculated the dominance metrics), you will want to visualize the data. Here, I will first demonstrate how to visualize the stability of dominance ranks across time, since most of my publications focused on dominance stability. Then I will walk through some simple visualizations for comparing the frequency of dominance behaviors. All the visualizations are done with R using the ‘ggplot’ package. I also use the ‘cowplot’ package for the theme and organizing the charts into panels, and the ‘see’ package for colors.
This file can serve as a pipeline for visualizing dominance data from your own study, just download this R markdown file and change the data.
library(ggplot2)
library(cowplot)
library(see)
library(dplyr)##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
library(tidyr)this is the same data from my [‘tube test data analysis tutorial’](#this is the same data from my ‘tube test data analysis tutorial’ & my ‘frequency and duration dominance data analysis tutorial’ (forthcoming)) & my ‘frequency and duration dominance data analysis tutorial’ (forthcoming)
#tube test data
ranks_per_week <- read.csv("data/ranks_per_week.csv")
ranks_per_cage <- read.csv("data/ranks_per_cage.csv")
ranks_per_cage <- ranks_per_cage %>%
select(animal_id, rank) %>%
rename(cage_rank = rank)
inventory1 <- read.csv("data/animal_inventory1.csv")
sex_id1 <- inventory1 %>%
select(animal_id, sex)
ranks_per_week <- left_join(ranks_per_week, sex_id1, by = "animal_id")
ranks_per_week <- left_join(ranks_per_week, ranks_per_cage, by = "animal_id")This jittered line graph nicely depicts the stability of the animals’ dominance ranks across time, and the accompanying histogram shows the animals that remained stable throughtout (i.e., horizontal lines). This graph appears messy by design. The general thinking in biomedical research is that since mice are inbred (i.e., genetically identical) and their cages are standardized (i.e., structurally identical) then all animals should have the same behaviors — genes x environment = phenotype. However, the social dominance hierarchies mice inevitably form within a cage indicates that the environment is different for each animal because they occupy a different dominance rank / social status. Moreover, the rankings are mostly unpredictable, especially for non-alpha ranked mice.
A similar graph was used in my publication from 2018 10.1038/s41598-018-24624-4-4
#subset males and females
males <- filter(ranks_per_week, sex == "Male")
females <- filter(ranks_per_week, sex == "Female")
male_plot <- ggplot(males, aes(day, rank, group = animal_id, color = as.factor(cage_rank), label = animal_id)) +
geom_line(linewidth = 1, aes(y=jitter(rank, .5))) +
scale_x_discrete(labels = c("d1" = 1, "d2" = 2, "d3" = 3), expand = c(0,0)) +
scale_y_reverse(breaks = c(1,2,3,4,5), labels = c("Alpha", "Beta", "Gamma", "Delta", "Epsilon"), expand = c(0,0)) +
scale_color_social() + guides(colour="none") +
ggtitle("Males") +
xlab("Week") + ylab("") +
theme_minimal_vgrid()
female_plot <- ggplot(females, aes(day, rank, group = animal_id, color = as.factor(cage_rank), label = animal_id)) +
geom_line(linewidth = 1, aes(y=jitter(rank, .5))) +
scale_x_discrete(labels = c("d1" = 1, "d2" = 2, "d3" = 3), expand = c(0,0)) +
scale_y_reverse(breaks = c(1,2,3,4,5), labels = c("Alpha", "Beta", "Gamma", "Delta", "Epsilon"), expand = c(0,0)) +
scale_color_social() + guides(colour="none") +
ggtitle("Females") +
xlab("Week") + ylab("") +
theme_minimal_vgrid()
#make stability dataset
stability <- ranks_per_week %>%
select(cage_id, animal_id, sex, cage_rank, day, rank) %>%
pivot_wider(names_from = day, values_from = rank) %>%
mutate(stability = case_when(
d1 == d2 & d2 == d3 ~ "stable",
d1 != d2 ~ "unstable",
d1 != d3 ~ "unstable",
d2 != d3 ~ "unstable"
)) %>%
filter(stability == "stable")
#make stability histograms of only stable animals
males <- stability %>%
filter(sex == "Male")
females <- stability %>%
filter(sex == "Female")
male_hist <- ggplot(males, aes(cage_rank)) +
geom_bar() +
geom_hline(yintercept = 1:7,
color = "white", linewidth = 1) +
scale_x_reverse() + coord_flip() +
xlab("") + ylab("")+
theme_nothing()
female_hist <- ggplot(females, aes(cage_rank)) +
geom_bar() +
geom_hline(yintercept = 1:7,
color = "white", linewidth = 1) +
scale_x_reverse() + coord_flip() +
theme_nothing()
#geom_hline, removes the need for an axis, just make sure that the number of lines (e.g. 1:7) is equal to the highest number of animals for a specific rank. In this case, there were 7 female alphas that remained stable.
#plot everything in panels
plot_grid(male_plot, male_hist, female_plot + theme(axis.text.y = element_blank()), female_hist, ncol = 4, align = "h", rel_widths = c(1.2, .2, 1, .2))
While the above graph informs us on the relative chaos of dominance rankings across time, it does little to help us understand what is occurring at a cage-level. To rectify this, we can use faceting to make a panel for each cage of animals.
A similar graph was used in my publication from 2023 10.3390/ani14010064
ggplot(ranks_per_week, aes(day, rank, group = animal_id, color = sex)) +
geom_line() +
scale_y_reverse(breaks = c(1,2,3,4,5), labels = c("Alpha", "Beta", "Gamma", "Delta", "Epsilon")) +
scale_color_social() +
facet_wrap(~cage_id) + theme_cowplot()
In this case, I would manually go through each graph and begin assigning it a stability category to better understand the general stability between the sexes.
stability_categories <- data.frame(
cage_id = c(38, 41, 47, 50, 74, 77, 86, 110, 113, 119, 122, 146, 149, 155, 158, 173, 176, 182, 185),
stab_cat = c("unstable alpha", "unstable alpha", "double despotic", "despotic", "unstable alpha", "stable transitive", "despotic", "despotic", "despotic", "unstable alpha", "double despotic", "despotic", "despotic", "despotic", "despotic", "double despotic", "stable transitive", "despotic", "despotic")
)
ranks_per_week <- left_join(ranks_per_week, stability_categories, by = "cage_id")
ggplot(ranks_per_week, aes(day, rank, group = animal_id, color = stab_cat)) +
geom_line() +
scale_y_reverse(breaks = c(1,2,3,4,5), labels = c("Alpha", "Beta", "Gamma", "Delta", "Epsilon")) +
scale_color_social() +
labs(color = "Stability \nCategory") +
facet_wrap(~cage_id) + theme_cowplot()
We can also use David’s score to better understand the dominance ranks and structures. Briefly, David’s score measures the proportion of winning. This places the rankings on a continuous scale rather than a discrete scale, which is better for understanding animals that frequently switch ranks.
#import data
ds_scores <- read.csv("data/ds_per_cage.csv")
ds_scores <- left_join(ds_scores, sex_id1, by = "animal_id")
ds_scores <- left_join(ds_scores, stability_categories, by = "cage_id")
ds_scores <- left_join(ds_scores, ranks_per_cage, by = "animal_id")
ggplot(ds_scores, aes(stab_cat, ds, fill = as.factor(cage_rank)))+
geom_dotplot(binaxis = 'y', stackdir = 'center',
stackratio=1.5, dotsize=1.2) +
xlab("") + ylab("David's Score") +
scale_fill_social() + labs(fill = "Dominance \nRank") + theme_cowplot()## Bin width defaults to 1/30 of the range of the data. Pick better value with
## `binwidth`.
We can also use facet_wrap again to better understand the data. But this time, we will set the x-axis as the cage_id and facet_wrap on the stability categories we set. Note, since cage_id is a number, we must set it as.factor to make it a discrete variable rather than continuous. We can also use a vgrid theme to keep the animals in the same cage, visually.
ggplot(ds_scores, aes(as.factor(cage_id), ds, color = as.factor(cage_rank)))+
geom_point() +
xlab("") + ylab("David's Score") +
scale_color_social() + labs(color = "Dominance \nRank") +
facet_wrap(~stab_cat) +
theme_minimal_vgrid() + theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))