ggplot和boxplot：是否可以添加权重？

huangapple
go评论152阅读模式
英文:

ggplot & boxplot: is it possible to add weights?

问题

我正在尝试根据地区绘制工资的箱线图。

这是我的数据集的示例（由研究机构提供）：

  1. > head(final2, 20)
  2.    nquest nord ireg staciv etalav acontrib       nome_reg tpens   pesofit
  3. 1     173    1   18      3     25       35       Calabria  1800 0.3801668
  4. 2    2886    1   13      1     26       35        Abruzzo  1211 0.2383701
  5. 3    2886    2   13      1     20       42        Abruzzo  2100 0.2383701
  6. 4    5416    1    8      3     16       30 Emilia Romagna   700 0.8819879
  7. 5    7886    1    9      1     22       35        Toscana  2000 1.2452078
  8. 6   20297    1    5      1     14       39         Veneto  1200 1.6694498
  9. 7   20711    2    4      1     15       37       Trentino  2000 3.3746801
  10. 8   22169    1   15      4     40        5       Campania   600 1.6875562
  11. 9   22276    1    8      2     18       37 Emilia Romagna  1200 2.1782894
  12. 10  22286    1    8      1     15       19 Emilia Romagna   850 3.0333999
  13. 11  22286    2    8      1     15       35 Emilia Romagna   650 3.0333999
  14. 12  22657    1   16      1     25       40         Puglie  1400 0.3616937
  15. 13  22657    2   16      1     26       36         Puglie  1500 0.3616937
  16. 14  23490    1    5      2     23       36         Veneto  1400 0.9763965
  17. 15  24147    1    4      1     26       35       Trentino  1730 1.2479984
  18. 16  24147    2    4      1     18       45       Trentino  1600 1.2479984
  19. 17  24853    1   11      1     18       38         Marche  2180 0.3475683
  20. 18  27238    1   12      1     16       31          Lazio  1050 3.6358952
  21. 19  27730    1   20      1     15       37       Sardegna  1470 0.7232677
  22. 20  27734    1   20      1     16       45       Sardegna  1159 0.6959107

变量说明：

  1. nquest = 家庭代码
  2. nord = 家庭成员
  3. nome_reg = 他们所居住的地区
  4. tpens = 每个人赚的工资
  5. pesofit = 每个观测的权重

这是我正在使用的代码：

  1. final2 %>%
  2.   filter(nome_reg == "Piemonte"|
  3.          nome_reg == "Valle D'Aosta" | 
  4.          nome_reg == "Lombardia" | 
  5.          nome_reg == "Liguria"
  6.         ) %>%
  7.   ggplot(aes( x = factor(nome_reg, 
  8.                       levels=c("Piemonte", "Valle D'Aosta", "Lombardia", "Liguria")), 
  9.              y = tpens , fill = nome_reg ))+
  10.   geom_boxplot(varwidth = TRUE)

这给我这个图：

[ggplot和boxplot：是否可以添加权重？]

有没有办法绘制加权箱线图？也就是说，绘制一个考虑每个观察的权重（在这种情况下，每个地区中每个个体的工资 tpens ）的箱线图？

我已经在进行加权回归，因此我想可视化加权数据。

我尝试在 aes 中使用 weight = pesofit：

  1. final2 %>%
  2.   filter(nome_reg == "Piemonte"|
  3.          nome_reg == "Valle D'Aosta" | 
  4.          nome_reg == "Lombardia" | 
  5.          nome_reg == "Liguria") %>%
  6.   ggplot(aes( x = factor(nome_reg, levels=c("Piemonte", "Valle D'Aosta", "Lombardia", "Liguria")), 
  7.              y = tpens , fill = nome_reg, weight = pesofit ))+
  8.   geom_boxplot(varwidth = TRUE)

但R给出警告信息：

  1. Warning message:
  2. The following aesthetics were dropped during statistical transformation: weight
  3. i This can happen when ggplot fails to infer the correct grouping structure in the data.
  4. i Did you forget to specify a `group` aesthetic or to convert a numerical variable into a factor?

如何解决这个问题？

英文:

I'm trying to plot boxplots about wage according with the area.

This is a sample of my dataset ( It is provided by a research institute)

  1. > head(final2, 20)
  2. nquest nord ireg staciv etalav acontrib       nome_reg tpens   pesofit
  3. 1     173    1   18      3     25       35       Calabria  1800 0.3801668
  4. 2    2886    1   13      1     26       35        Abruzzo  1211 0.2383701
  5. 3    2886    2   13      1     20       42        Abruzzo  2100 0.2383701
  6. 4    5416    1    8      3     16       30 Emilia Romagna   700 0.8819879
  7. 5    7886    1    9      1     22       35        Toscana  2000 1.2452078
  8. 6   20297    1    5      1     14       39         Veneto  1200 1.6694498
  9. 7   20711    2    4      1     15       37       Trentino  2000 3.3746801
  10. 8   22169    1   15      4     40        5       Campania   600 1.6875562
  11. 9   22276    1    8      2     18       37 Emilia Romagna  1200 2.1782894
  12. 10  22286    1    8      1     15       19 Emilia Romagna   850 3.0333999
  13. 11  22286    2    8      1     15       35 Emilia Romagna   650 3.0333999
  14. 12  22657    1   16      1     25       40         Puglie  1400 0.3616937
  15. 13  22657    2   16      1     26       36         Puglie  1500 0.3616937
  16. 14  23490    1    5      2     23       36         Veneto  1400 0.9763965
  17. 15  24147    1    4      1     26       35       Trentino  1730 1.2479984
  18. 16  24147    2    4      1     18       45       Trentino  1600 1.2479984
  19. 17  24853    1   11      1     18       38         Marche  2180 0.3475683
  20. 18  27238    1   12      1     16       31          Lazio  1050 3.6358952
  21. 19  27730    1   20      1     15       37       Sardegna  1470 0.7232677
  22. 20  27734    1   20      1     16       45       Sardegna  1159 0.6959107

The variables:

  1. nquest = is the code of the family
  2. nord = is the component of the family
  3. nome_reg = is the area where they live
  4. tpens = is the wage that each one of them earn
  5. pesofit = is the weight for each observation

This is the code I'm using

  1. final2 %>%
  2. filter(nome_reg == "Piemonte"| 
  3. nome_reg == "Valle D'Aosta" | 
  4. nome_reg == "Lombardia" | 
  5. nome_reg == "Liguria"
  6. ) %>%
  7. ggplot(aes( x = factor(nome_reg, 
  8. levels=c("Piemonte", "Valle D'Aosta", "Lombardia", "Liguria")), 
  9. y = tpens , fill = nome_reg ))+
  10. geom_boxplot(varwidth = TRUE)

Which gives me this plot

ggplot和boxplot：是否可以添加权重？

Is there a way to plot a weighted boxplot?? I mean a boxplot that takes into account the weights for each observation ( in this case the wage tpens for each individual in each area)?

I'm already performing a weighted regression, hence I would like to visualize the weighted data

I've tried weight = pesofit in aes

  1.  final2 %>%
  2. filter(nome_reg == "Piemonte"| 
  3. nome_reg == "Valle D'Aosta" | 
  4. nome_reg == "Lombardia" | 
  5. nome_reg == "Liguria") %>%
  6. ggplot(aes( x = factor(nome_reg, levels=c("Piemonte", "Valle D'Aosta", "Lombardia", "Liguria")), 
  7. y = tpens , fill = nome_reg, weight = pesofit ))+
  8. geom_boxplot(varwidth = TRUE)

but R answers

  1. Warning message:
  2. The following aesthetics were dropped during statistical transformation: weight
  3. i This can happen when ggplot fails to infer the correct grouping structure in the data.
  4. i Did you forget to specify a `group` aesthetic or to convert a numerical variable into a factor?

How can I solve??

答案1

得分: 2

基于一个简单的例子，似乎指定权重会按预期的方式工作，尽管有警告，请参见以下权重如何影响绘图的简单示例：

  1. set.seed(0)
  2. tmp <- data.frame(x=rnorm(100))   #要绘制的一些随机数据
  3. tmp$y <- ifelse(tmp$x>0, 1, 0.1)  #将正值的权重设置得很高
  5. ggplot(tmp, aes(x=x)) + geom_boxplot()    
  7. [![没有权重的输出][1]][1]
  8.     
  9. ggplot(tmp, aes(x=x, weight=y)) + geom_boxplot()
  10. #警告信息：
  11. #在统计变换期间被省略的以下美学属性：weight
  12. #ℹ 当ggplot未能推断数据中的正确分组结构时，可能会发生这种情况。
  13. #ℹ 您是否忘记指定`group`美学属性，或将数值变量转换为因子？
  15. [![带有权重的输出][2]][2]
  17. 看起来这个警告可能是虚假的，可能与[这个错误][3]有关。
  19.   [1]: https://i.stack.imgur.com/q08fP.png
  20.   [2]: https://i.stack.imgur.com/lEvRx.png
  21.   [3]: https://github.com/tidyverse/ggplot2/issues/5053
英文:

Based on a simple example, it seems that specifying the weights does what's expected, despite the warning, see the following simple example of how the weights affect the plot:

  1. set.seed(0)
  2. tmp &lt;- data.frame(x=rnorm(100))   #Some random data to plot
  3. tmp$y &lt;- ifelse(tmp$x&gt;0, 1, 0.1)  #weight positive values highly
  4. ggplot(tmp, aes(x=x)) + geom_boxplot()

ggplot和boxplot：是否可以添加权重？

  1. ggplot(tmp, aes(x=x, weight=y)) + geom_boxplot()
  2. #Warning message:
  3. #The following aesthetics were dropped during statistical transformation: weight
  4. #ℹ This can happen when ggplot fails to infer the correct grouping structure in the data.
  5. #ℹ Did you forget to specify a `group` aesthetic or to convert a numerical variable into a factor?

ggplot和boxplot：是否可以添加权重？

It seems like the warning may be spurious, possibly related to this bug

huangapple
  • 本文由 发表于 2023年3月9日 17:06:18
  • 转载请务必保留本文链接：https://go.coder-hub.com/75682432.html
  • boxplot
  • dplyr
  • ggplot2
  • r
  • weighted-average
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定