Grouped scatter plots

Approaching grouped scatterplots one step at a time.

Libraries

library(tidyverse)
library(ggpubr) #to add statistical significance

Cut me the chase and take me to the final ggplot output.

We are working on ChickWeight dataset for today.

To know more about this dataset, type ?ChickWeight in your console and click enter.

We are using 3 variables from this dataset: Time: is the number of days since birth of the chick, weight: of the chick in grams and Diet: experimental diet on which the chick was fed. We would like to check

– To what extent time since birth and weight of the chick are correlated? – Does this vary based on different diet (esp. 1 and 3)?

We are about to use scatter plot for the first qn and color them based on diets 1 and 3 to find the answer for the second.

Base R method

# base scatter
plot(ChickWeight$Time, ChickWeight$weight,
     col=factor(ChickWeight$Diet),
     pch=16,
     main = "Relation between time since birth and weight of chicken")
# to add legends
legend("topleft",
       legend = levels(factor(ChickWeight$Diet)),
       pch = 16,
       col=factor(levels(factor(ChickWeight$Diet))))

I really cannot infer anything from this plot.

Lets get the easier ggplot way.

ggplot way

Initiate the scatterplot

ChickWeight %>% 
  ggplot(aes(x=Time, y=weight))+
  geom_point()

Seems like weight increases with increasing time since birth, of course, understandable.

Adding axes titles

ChickWeight %>% 
  ggplot(aes(x=Time, y=weight))+
  geom_point()+
  labs(x="Time since birth (in days)",
       y="Weight of the chick (in grams)")

How about adding a regression line to confirm the direction of association?

Add regression line

ChickWeight %>% 
  ggplot(aes(x=Time, y=weight))+
  geom_point()+
  geom_smooth(method = "lm")+
  labs(x="Time since birth (in days)",
       y="Weight of the chick (in grams)")+
  theme_classic()

To know about more types of regression lines, read method argument under ?geom_smooth

For now,adding R and a p notation will step-up the game. We achieve that using ggpubr

ChickWeight %>% 
  ggplot(aes(x=Time, y=weight))+
  geom_point()+
  geom_smooth(method = "lm")+
  ggpubr::stat_cor(r.digits = 3)+
  labs(x="Time since birth (in days)",
       y="Weight of the chick (in grams)")+
  theme_classic()

How about adding different colors for the chicken based on the diet? There are 4 diet available in this dataset, lets choose just 2 so that the plot is not crowded.

Grouped scatter plots

dietselect <- c(1,3) #i am choosing 1st and 3rd diet
ChickWeight %>% 
  filter(Diet %in% dietselect) %>% # this is a way to filter more than 1 variable
  ggplot(aes(x=Time, y=weight, color=Diet))+
  geom_point()+
  geom_smooth(method = "lm")+
  ggpubr::stat_cor(r.digits = 3)+
  labs(x="Time since birth (in days)",
       y="Weight of the chick (in grams)")+
    scale_color_manual(values=c("#4F788D", "#E87F4D"))+
  theme_classic()

We see that chicken on diet 3 gained more weight than those on diet 1.

Happy scattering until I see you with the next post!

---

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.