how create custom labels on a circular line chart

Question

I am trying to figure out how to plot data on weekly basis while having labels on circular axis for months only.

This topic is partly covered here but the data they have there is structurally different.

I have a dataset with number of posts posted per week:

   id week posts   weekdate year month  col fct_month
1:  1    1    26 2001-01-01 2001     1 grey       Jan
2:  2    2    24 2001-01-08 2001     1 grey       Jan
3:  3    3    35 2001-01-15 2001     1 grey       Jan
4:  4    4    35 2001-01-22 2001     1 grey       Jan
5:  5    5    19 2001-01-29 2001     1 grey       Jan
6:  6    6    55 2001-02-05 2001     2 grey       Feb

I need to create a circular plot for each year. I can do this with labels for each week:


library(data.table)
library(dplyr)
library(ggplot2)
library(hrbrthemes)

a<-"
id   week posts   weekdate year month  col fct_month
1     1    26 2001-01-01 2001     1 grey       Jan
2     2    24 2001-01-08 2001     1 grey       Jan
3     3    35 2001-01-15 2001     1 grey       Jan
4     4    35 2001-01-22 2001     1 grey       Jan
5     5    19 2001-01-29 2001     1 grey       Jan
6     6    55 2001-02-05 2001     2 grey       Feb
7     7    32 2001-02-12 2001     2 grey       Feb
8     8    31 2001-02-19 2001     2 grey       Feb
9     9    36 2001-02-26 2001     2 grey       Feb
10   10    36 2001-03-05 2001     3 grey       Mar
11   11    24 2001-03-12 2001     3 grey       Mar
12   12    42 2001-03-19 2001     3 grey       Mar
13   13    29 2001-03-26 2001     3 grey       Mar
14   14    57 2001-04-02 2001     4 grey       Apr
15   15    31 2001-04-09 2001     4 grey       Apr
16   16    33 2001-04-16 2001     4 grey       Apr
17   17    17 2001-04-23 2001     4 grey       Apr
18   18    16 2001-04-30 2001     4 grey       Apr
19   19    35 2001-05-07 2001     5 grey       May
20   20    23 2001-05-14 2001     5 grey       May
"
mindf<-fread(a)
ggplot(data=mindf, aes(x=factor(week), y=posts, group = year, color = col)) + 
  ylim(0,NA) +
  geom_point(, stat='identity') +
  geom_polygon(fill=NA)+
  coord_polar(start =0) +
  scale_color_identity() +
  xlab('Week of the year')+
  ylab('Job ads posted per week') +
  theme_ipsum()

and the result is:

but when I do it with labeling months it all fails because I guess it becomes grouped by months:

ggplot(mindf, aes(x = fct_month, y = posts, 
               group = year, colour = col)) + 
  geom_line(size = .5) +
  # geom_polygon(fill=NA)+
  scale_color_identity()+
  # scale_y_discrete(expand = c(0, 1), breaks = NULL) +
  xlim(month.abb) + 
  coord_polar() +
  theme_bw() + xlab(NULL) + ylab(NULL)

UPDATE:

If I add scale_x_date(date_breaks = "1 month", date_labels = "%b") as @stefan suggests below, that results in multiple months with the same name:

Here is the reproducible code for checking that:

library(data.table)
library(dplyr)
library(ggplot2)


a<-" week posts   weekdate year month  col fct_month
    1    26 2001-01-01 2001     1 grey       Jan
    2    24 2001-01-08 2001     1 grey       Jan
    3    35 2001-01-15 2001     1 grey       Jan
    4    35 2001-01-22 2001     1 grey       Jan
    5    19 2001-01-29 2001     1 grey       Jan
    6    55 2001-02-05 2001     2 grey       Feb
    7    32 2001-02-12 2001     2 grey       Feb
    8    31 2001-02-19 2001     2 grey       Feb
    9    36 2001-02-26 2001     2 grey       Feb
   10    36 2001-03-05 2001     3 grey       Mar
    2    34 2002-01-07 2002     1 grey       Jan
    3    34 2002-01-14 2002     1 grey       Jan
    4    27 2002-01-21 2002     1 grey       Jan
    5    27 2002-01-28 2002     1 grey       Jan
    6    25 2002-02-04 2002     2 grey       Feb
    7    36 2002-02-11 2002     2 grey       Feb
    8    34 2002-02-18 2002     2 grey       Feb
    9    41 2002-02-25 2002     2 grey       Feb
   10    34 2002-03-04 2002     3 grey       Mar
   11    34 2002-03-11 2002     3 grey       Mar
"
mindf<-fread(a) %>% 
  mutate(weekdate=as.Date(weekdate, format="%Y-%m-%d"))
ggplot(data=mindf, aes(x=weekdate, y=posts, group = year, color = col)) + 
  ylim(0,NA) +
  geom_point(stat='identity') +
  geom_polygon(fill=NA, aes(group=year,color=year))+
  coord_polar(start =0) +
  scale_color_identity() +
  xlab('Week of the year')+
  ylab('Job ads posted per week') +
  scale_x_date(date_breaks = "1 month", date_labels = "%b")

stefan · Answer 1 · 2020-12-23T13:11:15.080

1

If you want month labels you could use a date scale like so:

library(data.table)
library(hrbrthemes)
library(ggplot2)

mindf<-fread(a)
mindf$weekdate <- as.Date(mindf$weekdate)
ggplot(data=mindf, aes(x=weekdate, y=posts, group = year, color = col)) + 
  ylim(0,NA) +
  geom_point(, stat='identity') +
  geom_polygon(fill=NA)+
  coord_polar(start =0) +
  scale_color_identity() +
  scale_x_date(date_breaks = "1 month", date_labels = "%b") +
  xlab('Week of the year')+
  ylab('Job ads posted per week') +
  theme_ipsum()

EDIT In case of multiple years a date scale is still possible. In that case we could fix the year of the date mapped on x, which I call monthdate in my code:

library(data.table)
library(hrbrthemes)
library(ggplot2)

mindf<-fread(a)
mindf$weekdate <- as.Date(mindf$weekdate)
mindf$day <- format(mindf$weekdate, "%d")
mindf$monthdate <- as.Date(paste("2001", mindf$month, mindf$day, sep = "-"))

ggplot(data=mindf, aes(x=monthdate, y=posts, group = year, color = col)) + 
  ylim(0,NA) +
  geom_point(, stat='identity') +
  geom_polygon(fill=NA)+
  coord_polar(start =0) +
  scale_color_identity() +
  scale_x_date(date_breaks = "1 month", date_labels = "%b") +
  xlab('Week of the year')+
  ylab('Job ads posted per week') +
  theme_ipsum()

edited Dec 23 '20 at 13:11

answered Dec 23 '20 at 10:13

stefan

90,330
6
25
51

Thanks! but see update. Unfortunately in this case for several years it starts repeating month names (several januaries, februaries etc) – Philipp Chapkovski Dec 23 '20 at 12:48
here is an example: https://rnotebook.io/anon/98e4e50e6a102030/notebooks/circiular.ipynb – Philipp Chapkovski Dec 23 '20 at 13:03
Hi Philipp. Okay. I See. Have a look at my edit. Maybe this offers a solution for your issue. – stefan Dec 23 '20 at 13:11
So it basically substitute year in all data to a fixed one (2001)? – Philipp Chapkovski Dec 23 '20 at 13:13
1

Yep. As far as I get it this should work. Of course you can choose whatever year you want. (; – stefan Dec 23 '20 at 13:15
it works like charm, thanks a lot! Do you think it is possible to NOT connect endline and startline in polar coordinates? – Philipp Chapkovski Dec 23 '20 at 13:23
Forget it! It's enough to change `geom_polygon` to `geom_line` – Philipp Chapkovski Dec 23 '20 at 13:28
very minor remark: it's better to choose the leap year (2020 for instance) because otherwise this code will produce NAs in this case: `as.Date("2001-02-29", format('%Y-%m-%d'))` – Philipp Chapkovski Dec 23 '20 at 15:47
1

(: Yep. Had the same thought on my bike ride. One always forgets about the details when not working with the real dataset. Best S. – stefan Dec 23 '20 at 15:51

how create custom labels on a circular line chart

1 Answers1