{"id":12576,"date":"2019-12-27T11:18:11","date_gmt":"2019-12-27T16:18:11","guid":{"rendered":"https:\/\/rud.is\/b\/?p=12576"},"modified":"2019-12-28T07:00:58","modified_gmt":"2019-12-28T12:00:58","slug":"short-attention-span-theatre-reproducing-axios-1-big-thing-google-trends-2019-news-in-review-with-ggplot2","status":"publish","type":"post","link":"https:\/\/rud.is\/b\/2019\/12\/27\/short-attention-span-theatre-reproducing-axios-1-big-thing-google-trends-2019-news-in-review-with-ggplot2\/","title":{"rendered":"Short Attention Span Theatre: Reproducing Axios’ “1 Big Thing” Google Trends 2019 News In Review with {ggplot2}"},"content":{"rendered":"

I woke up to Axios’ “1 Big Thing”<\/a> ridgeline chart showing the crazy that was the 2019 news cycle:<\/p>\n

\"\"<\/a><\/p>\n

and, I decided to reproduce it in {ggplot2}.<\/p>\n

Getting The Data<\/h3>\n

First, I had to find the data. The Axios chart is interactive, so I assumed the visualization was built on-load. It was, but the data was embedded in a javascript file vs loaded as JSON via an XHR request<\/a>:<\/p>\n

\"\"<\/a><\/p>\n

which was easy enough to turn into JSON<\/a> anyone can use.<\/p>\n

NOTE: The # hrbrmstr\/hrbrthemes<\/code> is an indication you may need to use the version of {hrbrthemes} from my gitea\/sourcehut\/gitlab\/bitbucket\/github. That package has instructions for installing fonts needed. Sub out theme_ipsum_es()<\/code> with theme_ipsum()<\/code>, theme_ipsum_rc()<\/code> or just use theme_bw()<\/code> and tweak aesthetics manually.<\/p>\n

library(ggalt)\nlibrary(hrbrthemes) # hrbrmstr\/hrbrthemes\nlibrary(tidyverse)\n\njsonlite::fromJSON(\"https:\/\/rud.is\/dl\/2019-axios-news.json\") %>% \n  as_tibble() -> xdf\n\nxdf\n## # A tibble: 31 x 3\n##    name                    avg data      \n##    <chr>                 <dbl> <list>    \n##  1 Gov't shutdown        20.5  <int [51]>\n##  2 Mexico-U.S. border    22.8  <int [51]>\n##  3 Green New Deal        11.3  <int [51]>\n##  4 Blackface              9.61 <int [51]>\n##  5 N. Korea-Hanoi Summit 11.2  <int [51]>\n##  6 Boeing 737 Max         4.79 <int [51]>\n##  7 Brexit                28.5  <int [51]>\n##  8 Israel                42.1  <int [51]>\n##  9 SpaceX                24.1  <int [51]>\n## 10 Game of Thrones       16.8  <int [51]>\n## # \u2026 with 21 more rows\n<\/code><\/pre>\n
This is pretty tidy already, but we’ll need to expand the data<\/code> column and give each week an index:<\/p>\n
unnest(xdf, data) %>% \n  group_by(name) %>% \n  mutate(idx = 1:n()) %>% \n  ungroup() %>% \n  mutate(name = fct_inorder(name)) -> xdf # making a factor foe strip\/panel ordering \n\nxdf\n## # A tibble: 1,581 x 4\n##    name             avg  data   idx\n##    <fct>          <dbl> <int> <int>\n##  1 Gov't shutdown  20.5    69     1\n##  2 Gov't shutdown  20.5   100     2\n##  3 Gov't shutdown  20.5    96     3\n##  4 Gov't shutdown  20.5   100     4\n##  5 Gov't shutdown  20.5    19     5\n##  6 Gov't shutdown  20.5     9     6\n##  7 Gov't shutdown  20.5    17     7\n##  8 Gov't shutdown  20.5     3     8\n##  9 Gov't shutdown  20.5     2     9\n## 10 Gov't shutdown  20.5     1    10\n## # \u2026 with 1,571 more rows\n<\/code><\/pre>\n
We’ll take this opportunity to find the first week of each month (via rle()<\/code>) so we can have decent axis labels:<\/p>\n
# get index placement for each month axis label\nsprintf(\"2019-%02s-1\", 1:51) %>% \n  as.Date(format = \"%Y-%W-%w\") %>% \n  format(\"%b\") %>% \n  rle() -> mons\n\nmons\n## Run Length Encoding\n##   lengths: int [1:12] 4 4 4 5 4 4 5 4 5 4 ...\n##   values : chr [1:12] \"Jan\" \"Feb\" \"Mar\" \"Apr\" \"May\" \"Jun\" \"Jul\" \"Aug\" \"Sep\" \"Oct\" ...\n\nmonth_idx <- cumsum(mons$lengths)-3\n\nmonth_idx\n##  [1]  1  5  9 14 18 22 27 31 36 40 44 48\n<\/code><\/pre>\n
We’ve got all we need to make a {ggplot2} version of the chart. Here’s the plan:<\/p>\n