Confusion plot with white zeros off diagonal - r

Question

I really like this confusion plot, but I want the values where Freq is zero to be white.
gg %>% ggplot(aes(Prediction, Reference, fill= Freq)) +
geom_tile() + geom_text(aes(label=Freq)) +
scale_fill_gradient(low="#f8766d", high="#00ba38") +
labs(x = "Prediction",y = "Reference")
I tried filtering out the zero Freq, but it looks ugly:
gg %>% dplyr::filter(Freq != 0) %>%
ggplot(aes(Prediction, Reference, fill= Freq)) +
geom_tile() + geom_text(aes(label=Freq)) +
scale_fill_gradient(low="#f8766d", high="#00ba38") +
labs(x = "Prediction",y = "Reference")
Anyone know how I can keep the red/green on the diagonal/off diagonal, but make zeros white?
Here is the data:
gg <- structure(list(Prediction = structure(c(1L, 2L, 3L, 4L, 5L, 6L,
7L, 8L, 9L, 10L, 1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 1L, 2L, 3L, 4L, 5L, 6L,
7L, 8L, 9L, 10L, 1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 1L, 2L, 3L, 4L, 5L, 6L,
7L, 8L, 9L, 10L, 1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 1L, 2L, 3L, 4L, 5L, 6L,
7L, 8L, 9L, 10L), .Label = c("0", "1", "2", "3", "4", "5", "6",
"7", "8", "9"), class = "factor"), Reference = structure(c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L,
2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 4L, 4L, 4L,
4L, 4L, 4L, 4L, 4L, 4L, 4L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L,
5L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 7L, 7L, 7L, 7L, 7L,
7L, 7L, 7L, 7L, 7L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 9L,
9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 10L, 10L, 10L, 10L, 10L,
10L, 10L, 10L, 10L, 10L), .Label = c("0", "1", "2", "3", "4",
"5", "6", "7", "8", "9"), class = "factor"), Freq = c(99L, 0L,
1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 97L, 2L, 0L, 0L, 0L, 0L,
0L, 1L, 0L, 0L, 1L, 93L, 1L, 2L, 1L, 0L, 1L, 1L, 0L, 0L, 0L,
1L, 96L, 0L, 0L, 0L, 0L, 3L, 0L, 0L, 0L, 3L, 0L, 94L, 0L, 0L,
2L, 0L, 1L, 2L, 0L, 0L, 2L, 1L, 85L, 1L, 0L, 6L, 3L, 0L, 0L,
1L, 0L, 1L, 1L, 95L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 3L, 0L, 0L,
89L, 0L, 8L, 1L, 0L, 1L, 1L, 0L, 0L, 0L, 0L, 97L, 0L, 1L, 0L,
0L, 1L, 1L, 1L, 0L, 2L, 2L, 92L)), class = "data.frame", row.names = c(NA,
-100L))

Answer 1

That was nice! I'm facing similar problems these days. If you want the 0 to appear, I think there are at least two ways of doing this:
The first one is by manipulating a bit the data and using the parameter na.value="white":
gg %>% mutate(Freq2 = ifelse(Freq == 0,NA,Freq)) %>%
ggplot(aes(Prediction, Reference, fill = Freq2)) +
geom_tile() +
geom_text(aes(label=Freq)) +
scale_fill_gradientn(colours = c("#f8766d", "#00ba38"),na.value="white") +
labs(x = "Prediction",y = "Reference", fill = "Freq")
The second one is by playing with the colours of the scale:
gg %>%
ggplot(aes(Prediction, Reference, fill = Freq)) +
geom_tile() +
geom_text(aes(label=Freq)) +
scale_fill_gradientn(colours = c("white","#f8766d", "#00ba38"), values = c(0,0.01,1)) +
labs(x = "Prediction",y = "Reference")
This solution has the downside of being quite sensible to the values, in the sense that if your data changes a lot, this won't lead to the same coloring. So the first solution is more robust.
To remove the 0's from the white cells, the best option is to set the label to an empty string in those cases. For exemple, using the first solution:
gg %>% mutate(Freq2 = ifelse(Freq == 0,NA,Freq)) %>%
ggplot(aes(Prediction, Reference, fill = Freq2)) +
geom_tile() +
geom_text(aes(label=ifelse(Freq == 0,"",Freq))) +
scale_fill_gradientn(colours = c("#f8766d", "#00ba38"),na.value="white") +
labs(x = "Prediction",y = "Reference", fill = "Freq")

Answer 2

One option is to use na.value after replacing the 0 to NA
library(ggplot2)
library(dplyr)
gg %>%
mutate(Freq = na_if(Freq, 0)) %>%
ggplot(aes(Prediction, Reference, fill= Freq)) +
geom_tile() +
geom_text(aes(label=Freq)) +
scale_fill_gradient(low="#f8766d", high="#00ba38",
na.value = 'white') +
labs(x = "Prediction",y = "Reference")
-output
If we need the 0 values as well, instead of creating a new column, can change the 0 to NA in fill using .data
ggplot(gg, aes(Prediction, Reference, fill = na_if(.data[["Freq"]], 0))) +
geom_tile() +
geom_text(aes(label=Freq)) +
scale_fill_gradient(low="#f8766d", high="#00ba38",
na.value = 'white') +
labs(x = "Prediction",y = "Reference", fill = "Freq")