如何在 R 中创建面板数据的多方面图？

Question

I would like to create a multi-facet graph based on panel data. While graphing panel data with only one y-variable is relatively simple, I am asking how can I create a graph in R using multiple y-variables that should appear in the same graph. The problem is that I have two "y". Each ggplot has the (...aes(x=year, y=something, ...) but I have two "y's", namely source1 and source2 and I can't find a solution to create a multi-facet graph that contains both y-variables in the same facet. Please see the description of the panel data below. The panel data that I want to graph in R looks like this:

structure(list(id = c(46L, 46L, 46L, 113L, 113L, 113L, 238L, 
238L, 238L, 2224L, 2224L, 2224L, 5557L, 5557L, 5557L, 757L, 757L, 
757L, 8890L, 8890L, 8890L, 33335L, 33335L, 33335L, 48L, 48L, 
48L, 115L, 115L, 115L, 240L, 240L, 240L, 2226L, 2226L, 2226L, 
5559L, 5559L, 5559L, 1478L), area = structure(c(1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L), .Label = c("Australia and New Zealand", "Brazil", 
"Canada", "China", "India", "United States of America"), class = "factor"), 

I've created the panel data in R:

panel <- pdata.frame(data, index = c("id", "year"), drop.index = FALSE)

Then I've tried this plot in ggplot:

    geom_area() +
    scale_fill_viridis(discrete = TRUE) +
    theme(legend.position="none") +
    ggtitle("Yield") +
    theme_ipsum() +
    theme(
      legend.position="none",
      panel.spacing = unit(0.1, "lines"),
      strip.text.x = element_text(size = 8),
      plot.title = element_text(size=14)
    ) +
    facet_wrap(~item)

however, it does not shows facets on area and then inside each facet the values of source1 and source2 values for each crop. I would like to create something like this:

The problem becomes even more complicate because I have many dimensions: area, item, years and those two y-variables: source1 and source2. The most important is to create facets that show source1 and source2 appear as lines or bars for comparison reasons. But the problem is how to create facets that show this two y-variables by year, area and item? And all of these without producing an overcrowded graph.

Because dc27 asked examples of graph, another possible example would be:

where source1 and source2 should be 2 bars side by side showing values per year, per item and per area. If you have any other suggestions on how to plot this panel data you are more than welcome.

Answer 1

Ok based on your question and discussion, you want to provide values for average and GLOBIOM as y axis but always been plotted side-by-side in order to compare various item and area.

Here, a possible way is to pivot your columns of interest for y in a longer format using for example pivot_longer function:

library(tidyr)
library(dplyr)
library(ggplot2)

data %>% pivot_longer(cols = c(average, GLOBIOM), names_to = "Var", values_to = "val")

# A tibble: 80 x 10
      id area                      element item    year value   diff relat        Var       val
   <int> <fct>                     <fct>   <fct>  <int> <int>  <dbl> <fct>        <chr>   <dbl>
 1    46 Australia and New Zealand Yield   Barley  2000 20080  0.35  17.32673267  average  2.02
 2    46 Australia and New Zealand Yield   Barley  2000 20080  0.35  17.32673267  GLOBIOM  1.67
 3    46 Australia and New Zealand Yield   Barley  2010 18268  0.15  7.894736842  average  1.9 
 4    46 Australia and New Zealand Yield   Barley  2010 18268  0.15  7.894736842  GLOBIOM  1.75
 5    46 Australia and New Zealand Yield   Barley  2018 23044  0.6   24.3902439   average  2.46
 6    46 Australia and New Zealand Yield   Barley  2018 23044  0.6   24.3902439   GLOBIOM  1.86
 7   113 Australia and New Zealand Yield   Maize   2000 58727 -0.93  -15.6302521  average  5.95
 8   113 Australia and New Zealand Yield   Maize   2000 58727 -0.93  -15.6302521  GLOBIOM  6.88
 9   113 Australia and New Zealand Yield   Maize   2010 67515 -0.570 -8.177905308 average  6.97
10   113 Australia and New Zealand Yield   Maize   2010 67515 -0.570 -8.177905308 GLOBIOM  7.54
# … with 70 more rows

Then, you can use "val" for y axis and "var" as fill argument for dodged barchart. Using facet_grid , you can separate data for various area and items.

Altogether, you can do something like this:

library(tidyr)
library(dplyr)
library(ggplot2)

data %>% pivot_longer(cols = c(average, GLOBIOM), names_to = "Var", values_to = "val") %>%
  ggplot(aes(x = element, y = val, fill = Var))+
  geom_col(position = position_dodge())+
  facet_grid(area~item, scales = "free")

Does it answer your question?

---

On your question, you mention to have plot being display per year, per item and per area. One way of doing this is to do:

data %>% pivot_longer(cols = c(average, GLOBIOM), names_to = "Var", values_to = "val") %>%
  ggplot(aes(x = as.factor(year), y = val, fill = Var))+
  geom_col(position = position_dodge())+
  facet_grid(area~item, scales = "free")+
  theme(axis.text.x = element_text(angle = 60, hjust =1))