In k-hench/geomfactory: Parallel ggplot color scales

knitr::opts_chunk$set(echo = TRUE)

About the package:

The geomfactory package is evil in that it provides infrastructure to implement several parallel color scales within a single ggplot. This is usually not possible for good reasons, but I encountered some situations where I needed this feature:

Usually I use this approach in situations where I want to plot unrelated statistics that share the same spatial context. An example would be a population genetic measures that are all measured for the same reference genome. Here I might want to keep different color schemes for consistency within my project, yet still plot them together to highlight outlier regions in one measure.

The aim is NOT to overlay several unrelated data sets, but to use the ggplot facet functions to put them in the same context:

library(geomfactory)
ls(as.environment("package:geomfactory"))

After loading the package, currently there are only two new functions: factory_geom_line() & factory_geom_point(). This are our geom_factorys that will create new geom_line() (or geom_point()) variations that will work on a alternative color scale. The function also creates according scales scale_*() (one discrete, one continuous) as well as the needed color-bar guide guide_colourbar_*().

(This might expand in the future to also include other geoms.)

factory_geom_line('var2')
ls(.GlobalEnv)

We can do this for as may additional color scales as we need (using a different name for each scale).

factory_geom_point('var3')
ls(.GlobalEnv)

Example

Now let see the new color scales in action:

First, we create three data sets that all refer to the same spatial system (in this case our simple x axis), but which are otherwise unrelated:

set.seed(2)

data1 <- tibble(x = seq(-2*pi, 2*pi, length.out = 100),
       A = sin(x),
       B = cos(x),
       C = -sin(x),
       D = -cos(x)) %>% 
  gather(key = 'var1', value = 'y', 2:5) %>%
  mutate(var = 'var1')

These data sets include:

a x value
a y value
a varX vale or class (that's what we want to color)
a var column that is uniform within each data set (for the faceting)

data1 %>%
  DT::datatable(options = list(pageLength = 5))

data2<- tibble(x = seq(-2*pi, 2*pi, length.out = 100),
                `0` = x*1,
                `1` = x*2,
                `5` = x*4,
                `10` = x*10) %>% 
  gather(key = 'var2', value = 'y', 2:5, convert = TRUE) %>%
  mutate(var = 'var2')

data3 <- tibble(x = seq(-2*pi, 2*pi, length.out = 100),
                a = 1+rnorm(length(x)),
                b = 2+rnorm(length(x)),
                c = 5+rnorm(length(x))) %>% 
  gather(key = 'var3',value = 'y',2:4) %>%
  mutate(var = 'var3')

Finally, we can plot the three data sets in context while keeping their super important color scheme. We can also highlight the crazy outlier of the class c within var3:

clr_1 <- viridis::viridis(4)
clr_3 <- RColorBrewer::brewer.pal(3,'Set1')

ggplot(data1, aes(x,y))+
  geom_vline(xintercept = -1.08, color = rgb(0,0,0,.5))+
  geom_point(aes(color = var1))+
  geom_line_var2(data = data2, aes(var2 = var2, group = var2))+
  geom_point_var3(data = data3, aes(var3_f = var3), 
                  shape = 21, size = 2.5,alpha = .8)+
  scale_color_manual(values = clr_1)+
  scale_var3_f_manual(values = clr_3) +
  guides(var2 = guide_colourbar_var2(barheight = unit(9,'pt')),
         var3_f = guide_legend(override.aes = list(fill = clr_3)))+
  facet_grid(var~.,scales = 'free')+
  theme(legend.position = 'bottom')