5

我希望这个例子会很清楚。我想在中间条跨越“0”的地方堆叠条，因为它代表一个中性值。这与李克特量表一起使用。为了重现性，我使用了 diamonds 数据集。

下面的示例与我的用例足够接近，并演示了我很难让“好”或“正面”数据以正确的顺序排列（因此中性最接近 0）。

这是我的代码：

require(tidyverse)

diamonds_new <- diamonds %>%
  mutate(quality = fct_recode(cut, "Very poor" = "Fair", "Poor" = "Good", "Neutral" = "Very Good", "Good" = "Premium", "Excellent" = "Ideal")) %>% 
  select(color, clarity, quality) %>% 
  group_by(color, clarity, quality) %>% count()

diamonds_bad <- 
  diamonds_new %>% filter(quality %in% c("Very poor", "Poor", "Neutral")) %>% 
  mutate(n = ifelse(quality == "Neutral", -n/2, -n))

diamonds_good <- 
  diamonds_new %>% filter(quality %in% c("Neutral", "Good", "Excellent")) %>% 
  mutate(n = ifelse(quality == "Neutral", n/2, n)) # %>% 
#  arrange(color, clarity, desc(quality))  # this doesn't seem to make a difference

ggplot() + geom_col(data = diamonds_bad, aes(x=color, y = n, fill = quality)) +  
  geom_col(data = diamonds_good, aes(x=color, y = n, fill = quality)) + 
  facet_grid(. ~ clarity, scales = "free") + 
  coord_flip()

在此处输入图像描述 我也尝试过使用scale_fill_manual()，但也没有找到一种方法。

我相信这比一些现有的例子更复杂，这些例子没有负值或需要的复杂性span 0。使用当前版本的 ggplot，我错过了什么？

另外，我是否正确地认为需要拆分正负集，或者至少这样做更容易？

4

2 回答 2

3

创建的列geom_col使用position_stackwhich 分别堆叠正值和负值，其中正值向上堆叠，负值向下堆叠。在此示例中，中心组Neutral通过将其设置为等于其原始值的一半，然后将其绘制为正值和负值来跨越 0。此外，对于正值，组的顺序需要颠倒。

这种方法有助于展示我使用的一些调查的结果，因此我将其制成一个函数以使其更通用。

library(tidyverse)
#
# summarize groups and save counts in variable quality_cnt
#
  diamonds_cnt <- diamonds %>%
    mutate(quality = fct_recode(cut, "Very_Poor" = "Fair", "Poor" = "Good",
                                "Neutral" = "Very Good", "Good" = "Premium", "Excellent" = "Ideal")) %>%
    select(color, clarity, quality) %>%
    group_by(color, clarity, quality) %>% summarize(quality_cnt = n())

# make function to plot counts    

  plot_ratings <- function(survey, rated_item, rating_cnt, rating, rating_cat, facet = "wrap") {
#
#  Input:   
#         rated_item  =  unquoted variable name of rated items
#         rating = unquoted variable name of ratings for each rated_items; 
#                  variable should be a factor ordered from lowest to highest 
#         rating_cnt = unquoted variable name of counts or frequencies for each rated_item 
#         rated_cat = unquoted variable name of categories of rated items
#         facet  = "grid" for all panels on one row or 
#                   "wrap" to spread panels across multiple rows
#
#  make arguments quosures
#
    rated_item <- enquo(rated_item)
    rating_cnt <- enquo(rating_cnt)  
    rating <- enquo(rating)
    rating_cat <- enquo(rating_cat)
#
# If number of rating levels is odd, find middle rating
#
  rating_levels <- levels(pull(survey, !!rating))
  mid_level <-  ceiling(length(rating_levels)/2)
  mid_rating <- ifelse(length(rating_levels)%%2 == 1, rating_levels[mid_level], NA_character_)  
#
# make local variabels for use with aes
# plot positive and negative columns separately
#
  survey <- survey %>% mutate( rating_plt = !!rating, rating_cnt_plt = !!rating_cnt)

  sp <- ggplot(survey, aes_(x = rated_item,  fill = rating)) + 
        geom_col(data=filter(survey, !!rating %in% tail(rating_levels, mid_level)),
                 aes( y = ifelse(rating_plt == mid_rating, .5*rating_cnt_plt, rating_cnt_plt)),
                 position = position_stack(reverse = TRUE )) +
        geom_col(data=filter(survey, !!rating %in% head(rating_levels, mid_level)),
                 aes( y = ifelse(rating_plt == mid_rating, -.5*rating_cnt_plt, -rating_cnt_plt)),
                 position = "stack") +
        labs(y = rating_cnt) +
        scale_fill_brewer(palette = "RdYlGn", direction = -1) +
        coord_flip() +
        switch(facet,
               grid = facet_grid( facets=rating_cat, scales = "free_x"),
               wrap = facet_wrap( facets=rating_cat, scales = "free_x"))
  plot(sp)
  } 
#
#  Use function to make charts
#
  plot_ratings(diamonds_cnt,  rated_item = color, rating_cnt = quality_cnt, 
               rating = quality, rating_cat = clarity, facet = "wrap")

这给出了图表

在此处输入图像描述

于 2018-07-06T21:44:04.603 回答
0

像这样的东西 - 我的关键变化是远离你可以自由控制开始和结束的地方geom_col。geom_rectangle

diamonds_new <-  diamonds %>%
  mutate(quality = fct_recode(cut, "Very poor" = "Fair", "Poor" = "Good", "Neutral" =     "Very Good", "Good" = "Premium", "Excellent" = "Ideal")) %>% 
  select(color, clarity, quality) %>% 
  group_by(color, clarity, quality) %>% 
  count() %>% 
  group_by(color, clarity) %>% 
  arrange(quality) %>%
  mutate(end = cumsum(n)) %>%  
  mutate(start = end-n) %>%
  mutate(offset = (end[quality=="Neutral"] + start[quality=="Neutral"])/2) %>%   
  mutate(start = start - offset,
         end = end - offset) %>%
  mutate(colStart = as.numeric(color) + 0.25,
         colEnd = as.numeric(color) + 0.75)

看到第二个答案（并没有看到任何来自 OP 的输入），我还包括了一个替代方面。

ggplot() + 
    geom_rect(data = diamonds_new, aes(xmin=colStart, xmax=colEnd, ymin=start, ymax=end, fill = quality)) +  
    facet_wrap(. ~ clarity, scales="free_x") +        
    coord_flip()

WaltS 的答案保留了 y 轴上的因子水平，这肯定更接近原始问题。然而，它需要对数据进行大量更改，所以我想保留我的替代答案是有价值的。

在此处输入图像描述

如果零需要在面板中居中，则需要进行xlim相应调整。

于 2018-07-06T07:02:24.720 回答