使用 addmargin 计算表中的加权边际平均值

Question

我创建了一个表格，其中包含人们参与按国家和性别划分的调查的平均天数。

   tbl_tmp <- df_tmp1 %>% 
                 group_by(country, Sex) %>%      
                 summarize(Mean_Days_Used = round(mean(Days_Used), 1)) %>% 
                 spread(Sex, Mean_Days_Used) 

结果是

# A tibble: 10 x 4
# Groups:   country [10]
   country Female     Male       `No Response`
   <chr>   <drtn>     <drtn>     <drtn>       
 1 AU       23.3 days    NA days NA days      
 2 CA       60.8 days  46.5 days NA days      
 3 FR         NA days  17.0 days NA days      
 4 GB       52.0 days  62.8 days 31 days      
 5 HK         NA days 125.0 days NA days      
 6 IE      109.6 days 100.5 days NA days      
 7 JP         NA days  10.0 days NA days      
 8 NZ      118.0 days   8.0 days NA days      
 9 PL         NA days    NA days 27 days      
10 US       35.2 days  41.7 days 86 days 

我现在想为每一行和每一列添加边际平均值，但有两个转折：

1. 我想要一个加权平均值（按每个单元格中条目的比例加权），而不仅仅是一个简单的行或列平均值
2. 我需要处理 NA，当没有来自特定国家的特定性别的受访者时会出现这种情况

我通常会使用addmargins 来添加所有边际总和/平均值等，但这无法解释每个单元格中进入者的相对频率，此外，还会阻塞 NA。例如，如果新西兰有 10 名男性受访者和 20 名女性受访者，我希望边缘列的第 8 行为 (118 * 10 + 8 * 20 + 0 )/30 = 44.67。类似的论点适用于加权列均值。

我应该如何计算这些加权平均值？

在此致以诚挚的谢意

托马斯·飞利浦

Answer 1

您可以尝试使用：

library(dplyr)
library(tidyr)

df_tmp1 %>% 
  group_by(country, Sex) %>%      
  summarise(Mean_Days_Used = mean(Days_Used),
            n = n()) %>%
  mutate(total = (Mean_Days_Used * n)/sum(n)) %>%
  ungroup %>%
  select(-n, -Mean_Days_Used) %>%
  pivot_wider(names_from = Sex, values_from = total) %>%
  mutate(mean = rowMeans(.[-1], na.rm = TRUE)) -> df1


result <- rbind(df1, data.frame(country = 'Total', 
                     t(colMeans(df1[-1], na.rm = TRUE)), check.names = FALSE))

Answer 2

我有一个笨拙的答案，但肯定有一个更干净的方法

   Avg_by_Bucket <- df_tmp1 %>%
                       group_by(country,Sex)%>%
                       summarize(Mean_Days_Used = round(mean(Days_Used), 1))%>%
                       spread(Sex, Mean_Days_Used)
    
    Avg_by_country <- df_tmp1 %>%
                        group_by(country)%>%
                        summarize(Mean_Days_Used = round(mean(Days_Used), 1))
    
    Avg_by_Sex <- df_tmp1 %>%
                    group_by(Sex)%>%
                    summarize(Mean_days_used_by_sex = round(mean(Days_Used), 1))%>%
                    spread(Sex, Mean_days_used_by_sex)
    
    Overall_Avg <- df_tmp1 %>%
                      summarize(Mean_Days_Used = round(mean(Days_Used), 1))
    
df_tmp_a <- full_join(Avg_by_Bucket, Avg_by_country, by = 'country')
df_tmp_b <- bind_cols(Avg_by_Sex, Overall_Avg)

table_3  <- bind_rows(df_tmp_a, df_tmp_b)
table_3[dim(table_3)[1], 1] <- "Mean_Days_Used"
column_to_rownames(table_3, var = "country")

当应用于我的玩具数据集时

> df_tmp1
# A tibble: 11 x 4
   UserId country Sex         Days_Used
   <chr>  <chr>   <chr>           <int>
 1 A      AU      Female              1
 2 B      AU      Female              2
 3 C      AU      Female              3
 4 D      AU      Male                4
 5 E      AU      Male                5
 6 F      CA      Female              6
 7 G      CA      Male                7
 8 H      CA      Male                8
 9 I      CA      No Response         9
10 J      CA      No Response        10
11 K      CA      No Response        11

使用

生成

df_tmp1 <- tibble(
                  UserId = LETTERS[1:11],
                  country = c(rep("AU", 5), rep("CA", 6)),
                  Sex = c(rep("Female", 3), rep("Male", 2), "Female", rep("Male", 2), rep("No Response", 3) ),                              
                  Days_Used = 1:11
                  )

给我我想要的：

               Female Male No Response Mean_Days_Used
AU                  2  4.5          NA            3.0
CA                  6  7.5          10            8.5
Mean_Days_Used      3  6.0          10            6.0