Create a heatmap of ratio

Question

everyone! I am new to R and would like to create a heatmap. There is a data set with columns:

X: x coordinate
Y: y coordinate
Pet_type: type of pet (cat, dog, hamster, etc)
Owner_type: type of owner (adult male, adult female, kid)

Small dataset:

Owner	Pet	X	Y
Male	Dog	27.793	88.2128
Male	Hamster	37.7177	87.9776
Female	Cat	24.4547	87.3016
Kid	Cat	36.464	84.9169
Kid	Dog	29.4175	84.5433
Female	Lizard	37.9588	83.9029
Male	Guinea pig	44.8986	82.7822
Kid	Dog	26.6216	82.0757
Male	Hamster	46.2332	81.9817
Male	Cat	31.9716	81.7507
Female	Cat	22.8606	80.9761
Kid	Dog	29.744	80.7988
Kid	Lizard	32.2393	80.35
Female	Guinea pig	38.92	78.8604
Male	Dog	39.42	78.3604
Kid	Hamster	32.2632	87.8267

What would be the steps to create a heatmap which shows the ratio of one specific pet vs all pets in that specific bin? For example: I want to create a heatmap of Cats density and if the bin consists of 20 pets and 10 of those are cats - the bin's value is 0,5 or 50%, etc.

I am using ggplot and I got that far that I can see count of total pets in each bin. What manipulations should I do to the table, before feeding it to ggplot?

df %>% 
  ggplot(aes(X, Y))+
  geom_bin_2d(bins=15)

I am struggling to understand how to create a statement that I want to see Cats vs all pets ratio in all bins.

So yeah, I would really appreciate if someone could help me with this problem (probably an easy one)

Welcome to SO! Please consider posting a [reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example). — medium-dimensional, Nov 09 '22 at 13:40
In addition to try and provide a reproducible example, I think as a general rule the best way to start is to first calculate the values you want to plot (ratio of pets in bins), get the results in tidy format and then worry about plotting. — yoland, Nov 09 '22 at 13:56
@yoland I added a small dataset sample, what would be the steps needed before ggplot? — matissb, Nov 09 '22 at 17:29

Dan Adams · Answer 1 · 2022-11-10T15:10:11.863

There are a number of ways to do 2-D binning. One option is to can get {ggplot2} to make the bins for you and then normalize to the total count and re-plot it. Here you first built the plot using raw count and then pull out the calculated bins using ggplot2::ggplot_build() and do a standard group_by() %>% mutate(fract = x/sum(x). Then you can re-plot.

library(tidyverse)

n <- 5000

d <- tibble(x = rnorm(n),
            y = rnorm(n),
            pet = fct_infreq(sample(
              c("cat", "dog", "fish", "bird"), n, T, prob = c(4, 3, 2, 1)
            )))
p <- d %>%
  ggplot(aes(x, y)) +
  geom_bin_2d(aes(fill = after_stat(count))) +
  facet_wrap( ~ pet)

# original plot of count per animal per bin
p

# get underlying data
e <- ggplot_build(p)$data[[1]]

# normalize and then re-plot
e %>% 
  mutate(pet = fct_recode(PANEL, cat = "1", dog = "2", fish = "3", bird = "4")) %>% 
  group_by(xbin, ybin) %>% 
  mutate(fract = count/sum(count)) %>% 
  ggplot(aes(xmin, ymin)) +
  geom_tile(aes(fill = fract)) +
  facet_wrap(~pet)

^{Created on 2022-11-09 with reprex v2.0.2}

Create a heatmap of ratio

1 Answers1