Reorder factor not working on grouped data

Question

I have this dataframe that I call top_mesh_terms

structure(list(topic = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L), term = c("Diabetes Mellitus", 
"Depression", "Syndrome", "Diabetes Mellitus, Type 2", "Lung Diseases", 
"Colorectal Neoplasms", "Osteoarthritis", "Sclerosis", "Lymphoma", 
"Lung Diseases, Obstructive", "Diabetes Mellitus", "Disease", 
"Hypertension", "Syndrome", "Carcinoma", "Infection", "Coronary Disease", 
"Lung Neoplasms", "Obesity", "Infarction"), beta = c(0.0196989252285569, 
0.018472562347772, 0.0175512616261399, 0.0146680780420432, 0.0133507951269683, 
0.01224603797061, 0.0116799262133244, 0.0107893497000735, 0.00926496950657875, 
0.00891926541108836, 0.0324598963852768, 0.0198135918084849, 
0.0162689075944415, 0.0157166860189554, 0.014855885836076, 0.0127365678034364, 
0.0109544570325732, 0.00964124158432716, 0.00956596829604797, 
0.00880281359338067)), class = c("tbl_df", "tbl", "data.frame"
), row.names = c(NA, -20L))

As the title suggests, I'd like to reorder the term column by beta and then draw a bar chart. I hoped to see a bar chart with ordered bars, but that is not the case. Here is the code I used and the resulting graph:

top_mesh_terms %>% 
  group_by(topic) %>% 
  mutate(term = fct_reorder(term, beta)) %>%
  ungroup() %>% 
  ggplot(aes(term, beta)) +
  geom_bar(stat = "identity") +
  facet_wrap(~ topic, scales = "free") +
  coord_flip() +
  scale_y_continuous(labels = scales::percent_format()) +
  labs(x = "MeSH Term",
       y = "Beta")

Did you try https://stackoverflow.com/questions/18624394/ggplot-bar-plot-with-facet-dependent-order-of-categories or https://stackoverflow.com/questions/43176546/ggplot2-reorder-bars-from-highest-to-lowest-in-each-facet ? — Ronak Shah, Dec 10 '18 at 01:46

Gregor Thomas · Accepted Answer · 2018-12-10T05:14:53.957

Your problem is the group_by. The levels of a factor have a single order, it can't differ by group. If we get rid of your group_by and ungroup commands, everything works:

    top_mesh_terms %>% 
      mutate(term = reorder(term, beta)) %>%
      ggplot(aes(term, beta)) +
      geom_bar(stat = "identity") +
      facet_wrap(~ topic, scales = "free") +
      coord_flip() +
      scale_y_continuous(labels = scales::percent_format()) +
      labs(x = "MeSH Term",
           y = "Beta")

(Btw, forcats has some really nice functions, but if the only one you need is fct_reorder you might as well use base::reorder - it does the same thing without the extra package dependency.)

score 1 · Answer 2 · edited Dec 10 '18 at 05:13

1

How about this?

top_mesh_terms %>% 
  group_by(topic) %>% 
  mutate(term = fct_reorder(term, beta)) %>%
  ungroup() %>% 
  ggplot(aes(reorder(term, beta), beta)) +
  geom_bar(stat = "identity") +
  facet_wrap(~ topic, scales = "free") +
  coord_flip() +
  scale_y_continuous(labels = scales::percent_format()) +
  labs(x = "MeSH Term",
       y = "Beta")

I used the ggplot(aes(reorder(term, beta) to change the order.

edited Dec 10 '18 at 05:13

Gregor Thomas

136,190
20
167
294

answered Dec 10 '18 at 02:42

william3031

1,653
1
18
39

1

Sorry, I mistakenly added a footnote to your answer when I meant to add it to mine. Just rolled back the change. – Gregor Thomas Dec 10 '18 at 05:14

Reorder factor not working on grouped data

2 Answers2