geom_raster to visualize missing values with additional colorcode

Question

This question is a follow-up to my previous question: Adding color code (fill) to vis_miss plot

I would like to visualize the "missing info" in a data frame using geom_raster from ggplot2 in R while also highlighting some additional data structure using color-coding.

Solution attempt:

library(tidyverse)
x11()
airquality %>%
  mutate(id = row_number()) %>%
  gather(-c(id,Month), key = "key", value = "val") %>%
  mutate(isna = is.na(val)) %>%
  mutate(Month=as.factor(ifelse(isna==TRUE,NA,Month)))  %>%
  ggplot(aes(key, id, fill = Month)) +
    geom_raster() +
    labs(x = "Variable",
           y = "Row Number", title = "Missing values in rows") +
    coord_flip()

plot

This is almost what I want, but it would be nicer to separate the month and NA legends. Is that possible? (Note that my system does not allow me to use transparency (alpha)).

Are you looking for something like [this](https://stackoverflow.com/questions/52060601/ggplot-multiple-legends-arrangement) ? — Orlando Sabogal, Sep 16 '19 at 19:04

score 1 · Accepted Answer · edited Jun 20 '20 at 09:12

Here, I removed the legend for NA. If this doesn't serve your purpose properly, I can think of a hacky solution to add another legend for data vs. missing.

library(tidyverse)

airquality %>%
  mutate(id = row_number()) %>%
  gather(-c(id,Month), key = "key", value = "val") %>%
  mutate(isna = is.na(val)) %>%
  mutate(Month_Dummy=as.factor(ifelse(isna==TRUE,NA,Month)))  %>%
  mutate(Month=as.factor(Month))  %>% 
  ggplot() +
  geom_raster(aes(key, id, fill = Month)) +
  geom_raster(aes(key, id, fill = Month_Dummy)) +
  labs(x = "Variable",
       y = "Row Number", title = "Missing values in rows") +
  coord_flip()

Update:

The hacky solution that I can think of is adding a geom_point for just one of the missing and used that for the legend of missing data points. It's not the best in terms of appearance, but is the only solution I can think of.

library(tidyverse)

airquality %>%
  mutate(id = row_number()) %>%
  gather(-c(id,Month), key = "key", value = "val") %>%
  mutate(isna = is.na(val)) %>%
  mutate(Month_Dummy=as.factor(ifelse(isna==TRUE,NA,Month)))  %>%
  mutate(Month=as.factor(Month))  -> aqdf

ggplot(data = aqdf, aes(key, id)) +
  geom_raster(aes(fill = Month)) +
  geom_raster(aes(fill = Month_Dummy)) +
  geom_point(data=aqdf[aqdf$isna==TRUE,][1,], 
             aes(NA, id, colour = "NA"),
             inherit.aes = FALSE) +
  scale_color_manual(values=c("grey50")) +
  labs(x = "Variable", y = "Row Number", 
       title = "Missing values in rows", color = "Missing") +
  coord_flip() +
  theme(legend.key = element_rect(fill = "grey50"))

This is better than the previous, but I would prefer to have a second legend for the NA square, similar to what `vis_miss` in the `visdat` package does. Would you use two boxes for the missing/available data, and if so how would you color the available data? — user98563, Sep 16 '19 at 21:12
This is fine. I added a line with theme(legend.key = element_rect(fill = "grey50")) at the end to make it look nicer. — user98563, Sep 16 '19 at 21:49
@user98563 nice one. I edited and used `NA` instead of `key` in `geom_point` to not have any actual points graphed (if you look closely at the graph resulted by previous code, you'd have seen the point). I guess this now should completely meet your needs. Hope to see you around this community actively. Cheers. — M--, Sep 16 '19 at 21:57

geom_raster to visualize missing values with additional colorcode

1 Answers1

Update: