How to set the alpha based on a column?

Question

Consider this example:

library(dplyr)
library(sf)
library(tmap)

d <- data_frame(one = c(1,1,2,1,1,1,1),
                two = c(1,1,2,1,1,1,1))

std <- st_as_sf(d, coords = c('one', 'two'))

std %>% tm_shape() + tm_bubbles(alpha = 0.3)

You can see that point (1, 1) is darker because it appears 6 times in the data. Therefore, thanks to the alpha blending, these points add-up.

My problem is that I cannot store the dataset as it. What I have is only an aggregated version, like

d_agg <- d %>% group_by(one, two) %>% 
  summarize(count = n()) %>% 
  ungroup()

# A tibble: 2 x 3
    one   two count
  <dbl> <dbl> <int>
1     1     1     6
2     2     2     1

How can I reproduce the same exact chart as before, using d_agg and the corresponding count variable?

Of course, re-creating the initial dataframe above is not feasible solution because I have too many points (and some points are repeated too many times)

Just using:

std_agg %>% tm_shape() + tm_bubbles(col = 'count', alpha = 0.3)

does not work

You could recreate the original dataframe using `rep` rows n times? — zx8754, Jun 14 '18 at 19:42
I was thinking maybe you can assign `count == 1` with `grey100` then subtract `100` depending on `count` to create a darker color` e.g. `color <- ifelse(count == 1, "grey100", paste0("grey", (100 - count + 1)))` — Tung, Jun 14 '18 at 20:00
I don't know enough `sf/tmap` package to give a full answer. I am thinking, it must be possible take agg data row by row, expand then plot into variable (list), then keep appending following rows. At the end plot that list. This way we will only expand the data per group, not for all groups at once. — zx8754, Jun 14 '18 at 21:10
@Tung perhaps that is a good idea. However, count can take very large values as well) — ℕʘʘḆḽḘ, Jun 14 '18 at 21:13

score 3 · Accepted Answer · answered Jun 15 '18 at 07:38

3

Unfortunately, alpha is not (yet) an aesthetic, so it is not possible to do alpha = "count".

My question: do you really need alpha? Probably not if you do not use the color aesthetic. In that case, your approach to use color to model alpha transparency was actually good, but just needs a little configuration:

std_agg %>% tm_shape() + tm_bubbles(col = 'count', style = "cont", 
    palette =  "Greys", contrast = c(.3, .7), legend.col.show = FALSE)

answered Jun 15 '18 at 07:38

Martijn Tennekes

1,951
13
19

thanks! I wonder how can I use a palette like `viridis:magma` in this context? – ℕʘʘḆḽḘ Jun 15 '18 at 11:49
1

The github version of tmap (will soon be tmap 2.0 on CRAN) supports viridis palettes natively: `palette = "magma"`. For tmap 1.x, just use `palette = viridis:magma(10)` – Martijn Tennekes Jun 16 '18 at 13:10
thanks but in this context why 10 instead of 100 or 1? i dont understand how this is mapped to the data above – ℕʘʘḆḽḘ Jun 16 '18 at 13:11

www · Answer 2 · 2018-06-14T20:00:20.523

1

Here I show how to recreate your data frame d by using dplyr. Although it is not addressing your question about how to pass numeric value to the alpha argument in tm_bubbles, consider it as a workaround.

std_agg <- d_agg %>% 
  slice(rep(row_number(), times = count)) %>%
  st_as_sf(coords = c('one', 'two'))

std_agg %>% 
  tm_shape() + 
  tm_bubbles(alpha = 0.3)

In fact, this base R to expand the data frame is probably more intuitive.

d_agg[rep(1:nrow(d_agg), times = d_agg$count), ]

edited Jun 14 '18 at 20:00

answered Jun 14 '18 at 19:56

www

38,575
12
48
84

i appreciate but this cannot work because the resulting dataset would be too big – ℕʘʘḆḽḘ Jun 14 '18 at 19:59
1

@ℕʘʘḆḽḘ I know. This is just a workaround. Wait and see if some `tmap` experts can help. – www Jun 14 '18 at 20:01

How to set the alpha based on a column?

2 Answers2