One of the ways to represent before and after changes is using dumbbells
Lets jump into it.
We will work on the ChickWeight dataset for today. We will explore how much the chick weights changed from baseline to day 10 grouped by the 4 diets.
Lets create the dataframe of interest.
There are many time points of data available for weight, we choose day 0 and day 10 for simplicity.
We need the day 0 and day 10 weights in separate columns to make things easier, so we use pivot_wider from the tidyr which is a part of tidyverse library.
Cut me the chase and take me to the final plot
Create the dataframe
Filter
# this is how to filter only the two times of interest
ChickWeight %>%
filter(Time == 0 | Time == 10) %>%
rmarkdown::paged_table()# to avoid the entire dataset beimg displayed, i add pages
Note that it is not saved as a dataframe yet. I will describe the steps one-by-one.
pivot wider
ChickWeight %>%
filter(Time == 0 | Time == 10) %>% # after filtering, we group them
group_by(Diet, Time, Chick) %>%
pivot_wider(
id_cols = c(Diet, Chick),
names_from = Time,
values_from = weight, names_prefix = "time_"
) %>%
rmarkdown::paged_table()
Now we shall summarise the weight info and create the dataframe
summarise mean
df <- ChickWeight %>%
filter(Time == 0 | Time == 10) %>% # after filtering, we group them
group_by(Diet, Time, Chick) %>%
pivot_wider(
id_cols = c(Diet, Chick),
names_from = Time,
values_from = weight, names_prefix = "time_"
) %>%
mutate(wtdiff = time_10 - time_0) %>% # we add a difference column
group_by(Diet) %>%
summarise(
mean_0 = mean(time_0),
mean_10 = mean(time_10, na.rm = TRUE),
mean_diff = mean(wtdiff, na.rm = TRUE)
)
df
# A tibble: 4 × 4
Diet mean_0 mean_10 mean_diff
<fct> <dbl> <dbl> <dbl>
1 1 41.4 93.1 51.5
2 2 40.7 108. 67.8
3 3 40.8 117. 76.3
4 4 41 126 85 Now time for plots!
Initiate the dumbbells
For the dumbbells we need to specify what the x and xends are, here we need the baseline weight to be on the left and day 10 weight on the right.
ggplot(df) +
geom_dumbbell(
aes(
x = mean_0,
xend = mean_10,
y = Diet
)
)
We got the coordinates right, now time for some tweaking.
Improving the dumbbells
We are adding size and colors for the dumbbells
p <- ggplot(df) +
geom_dumbbell(
aes(
x = mean_0,
xend = mean_10,
y = Diet
),
# size and color of the connecting lines
size = 1, color = "#676875", alpha = 0.5,
# size and color of the left half of the dumbbell
size_x = 5, colour_x = "#F51F18",
# right side of the dumbbell
size_xend = 5, colour_xend = "#1E4CE8"
) +
theme_classic()
p
How about we add the legends?
Mark the legends
We need to mark the legends at only one place, so we will filter the data.
p <- p + geom_text(
data = filter(df, Diet == 4), # if you need the labels at diet 1, tweak accordingly
aes(
x = mean_0 + 6,
y = Diet,
label = "Baseline weight"
),
color = "#F51F18",
vjust = -1.5 # this is vertical justification for the labels
) +
geom_text(
data = filter(df, Diet == 4),
aes(
x = mean_10 - 8,
y = Diet,
label = "Weight on 10th day"
), color = "#1E4CE8",
vjust = -1.5
) +
theme(legend.position = "none")
p
Let’s up our game by adding the differences which we already have in our df
Add a column of mean difference
p <- p + geom_label(
data = df,
aes(
x = mean_10 + 7,
y = Diet,
label = str_glue("{round(mean_diff,0)} kg")
)
)
p
Lets add axes titles and captions
Final Plot
p + labs(
x = "Mean Weights of Chicken",
caption = "Difference in weights are given in text boxes"
)
Distill is a publication format for scientific and technical writing, native to the web.
Learn more about using Distill at https://rstudio.github.io/distill.