Altering thickness of line graph based on counts

Question

Dataframe "id" has the columns year, id, and matriline, where each row is an incident. I wanted to count the number of incidents by matriline per year, so I did:

events.bymatr = 
id %>%
group_by(year, matr, .drop = FALSE) %>%
dplyr::summarise(n = n()) %>%
ungroup()
events.bymatr

I plotted a line graph of the number of incidents over time, by matriline.

ggplot(events.bymatr, aes(x=year, y=n, group=matr)) + geom_line(aes(color=matr))

My question is twofold:

Is there a way I could recreate this line graph where the thickness of the lines is determined by how many IDs there were, per matriline? I imagine this would involve reshaping my data above but when I tried to group_by(year,matr,id,.drop=FALSE) my data came out all wonky.
I want to change the color palete so that each color is very distinct - how do I attach a new color palette? I tried using this c25 palette with this code but it makes all my lines disappear. ggplot(events.bymatr, aes(x=year, y=n, group=matr)) + geom_line(aes(color=c25))

Thanks so much in advance!

Output of "id" (shortened to just the first five rows per column):

> dput(id)
    structure(list(date = structure(c(8243, 8243, 8243, 8248, 8947,
    class = "Date"), year = c(1992L, 1992L, 1992L, 1992L, 1994L), 
    event.id = c(8L, 8L, 8L, 10L, 11L), id = structure(c(51L, 55L, 59L, 
    46L, 51L), .Label = c("J11", "J16", "J17", "J2", "J22"),
    class = "factor"), sex = structure(c(1L, 2L, 2L, 1L, 1L),
    .Label = c("0", "1"), class = "factor"), age = c(28L, 12L, 6L, 42L, 
    30L), matr = structure(c(20L, 20L, 20L, 11L, 20L), .Label = c("J2", 
    "J4", "J7", "J9", "K11"), class = "factor"),
    matralive = structure(c(2L, 2L, 2L, 2L, 2L),
    .Label = c("0", "1"), class = "factor"), pod = structure(c(3L, 3L, 
    3L, 3L, 3L), .Label = c("J", "K", "L"), class = "factor")),
    row.names = c(NA, -134L), class = c("tbl_df", "tbl", "data.frame"))

Output of events.bymatr:

> dput(events.bymatr)
      structure(list(year = c(1992L, 1992L, 1992L, 1992L, 1992L),
      matr = structure(c(1L, 2L, 3L, 4L, 5L), .Label = c("J2", "J4", 
      "J7", "J9", "K11"), class = "factor"), n = c(0L, 0L, 0L, 0L, 0L)), 
      row.names = c(NA, -380L), class = c("tbl_df", "tbl", 
      "data.frame"))

Can you include a simplified data set so we can run your code? That will reduce ambiguities and make it easier to help. — Jon Spring, Mar 01 '22 at 23:29
`geom_line` cannot change aesthetics mid-line; if you have a line that is one `group=` yet you want it to change colors/thickness/linetype mid-stroke, then you should switch to `geom_segment` — r2evans, Mar 01 '22 at 23:37
FYI, with over 14 unique lines, you are going to be hard-pressed to find any palette (color-alone) that will give you really good distinct color differences. I've found anecdotally that 6-8 is doable (depending on many factors), but many more than that and they are not as clear as one would want. — r2evans, Mar 01 '22 at 23:39

Jon Spring · Accepted Answer · 2022-03-02T20:43:52.590

As @r2evans noted, it is surprisingly hard to distinguish clearly among more than a handful of colors. I used an example 20-color scale here that does a pretty good job, but even so a few can be tricky to distinguish. Here's an attempt using the storms dataset included with dplyr.

library(dplyr)
storms %>%
  group_by(name, year) %>%
  summarize(n = n(), .groups = "drop") %>%   # = number of name per year View
  tidyr::complete(name, year = 1975:2015, fill = list(n = 0)) %>% 
  group_by(name) %>%
  mutate(total = sum(n)) %>%                 # = number of name overall
  ungroup()  %>% 
  filter(total %% 12 == 0) %>% # Arbitrary, to reduce scope of data for example
  ggplot(aes(year, n, color = name, size = total, group = name)) +
  geom_line() +
  guides(color = guide_legend(override.aes = list(size = 3))) +
  ggthemes::scale_color_tableau(palette = "Tableau 20")

Altering thickness of line graph based on counts

1 Answers1