分层与加权随机抽样

R 中的抽样

Richie Cotton

Data Evangelist at DataCamp

各国咖啡数

按原产国分组的咖啡豆。

top_counts <- coffee_ratings %>% 
  count(country_of_origin, sort = TRUE) %>% 
  head()

# A tibble: 6 x 2
  country_of_origin          n
  <chr>                  <int>
1 Mexico                   236
2 Colombia                 183
3 Guatemala                181
4 Brazil                   132
5 Taiwan                    75
6 United States (Hawaii)    73

¹ 为便于说明，数据集中将夏威夷和台湾作为国家列示，它们是重要的咖啡产区。

筛选 6 个国家

top_counted_countries <- c(
  "Mexico", "Colombia", "Guatemala",
  "Brazil", "Taiwan", "United States (Hawaii)"
)
coffee_ratings_top <- coffee_ratings %>% 
    filter(country_of_origin %in% top_counted_countries)

或，等价地

coffee_ratings_top <- coffee_ratings %>%
  semi_join(top_counts)

¹ 在"使用 dplyr 进行连接"第 3 章学习半连接（semi join）。

简单随机样本的计数

coffee_ratings_samp <- coffee_ratings_top %>% 
  slice_sample(prop = 0.1)

coffee_ratings_samp %>%
  count(country_of_origin, sort = TRUE) %>%
  mutate(percent = 100 * n / sum(n))

# A tibble: 6 x 3
  country_of_origin          n percent
  <chr>                  <int>   <dbl>
1 Guatemala                 24    27.3
2 Mexico                    23    26.1
3 Brazil                    12    13.6
4 Colombia                  11    12.5
5 Taiwan                     9    10.2
6 United States (Hawaii)     9    10.2

比较计数

总体

# A tibble: 6 x 3
  country_of_origin          n percent
  <chr>                  <int>   <dbl>
1 Mexico                   236   26.8 
2 Colombia                 183   20.8 
3 Guatemala                181   20.6 
4 Brazil                   132   15   
5 Taiwan                    75    8.52
6 United States (Hawaii)    73    8.30

10% 样本

# A tibble: 6 x 3
  country_of_origin          n percent
  <chr>                  <int>   <dbl>
1 Guatemala                 24    27.3
2 Mexico                    23    26.1
3 Brazil                    12    13.6
4 Colombia                  11    12.5
5 Taiwan                     9    10.2
6 United States (Hawaii)     9    10.2

按比例分层抽样

coffee_ratings_strat <- coffee_ratings_top %>%
  group_by(country_of_origin) %>% 
  slice_sample(prop = 0.1) %>%
  ungroup()

coffee_ratings_strat %>%
  count(country_of_origin, sort = TRUE) %>%
  mutate(percent = 100 * n / sum(n))

# A tibble: 6 x 3
  country_of_origin          n percent
  <chr>                  <int>   <dbl>
1 Mexico                    23   26.7 
2 Colombia                  18   20.9 
3 Guatemala                 18   20.9 
4 Brazil                    13   15.1 
5 Taiwan                     7    8.14
6 United States (Hawaii)     7    8.14

等量分层抽样

coffee_ratings_eq <- coffee_ratings_top %>%
  group_by(country_of_origin) %>% 
  slice_sample(n = 15) %>%
  ungroup()

coffee_ratings_eq %>%
  count(country_of_origin, sort = TRUE) %>%
  mutate(percent = 100 * n / sum(n))

# A tibble: 6 × 3
  country_of_origin          n percent
  <chr>                  <int>   <dbl>
1 Brazil                    15    16.7
2 Colombia                  15    16.7
3 Guatemala                 15    16.7
4 Mexico                    15    16.7
5 Taiwan                    15    16.7
6 United States (Hawaii)    15    16.7

加权随机抽样

coffee_ratings_weight <- coffee_ratings_top %>%
  mutate(
    weight = ifelse(country_of_origin == "Taiwan", 2, 1)
  ) %>% 
  slice_sample(prop = 0.1, weight_by = weight)

coffee_ratings_weight %>%
  count(country_of_origin, sort = TRUE) %>%
  mutate(percent = 100 * n / sum(n))

10% 加权样本

# A tibble: 6 x 3
  country_of_origin          n percent
  <chr>                  <int>   <dbl>
1 Mexico                    23   26.1 
2 Guatemala                 20   22.7 
3 Taiwan                    15   17.0 
4 Brazil                    12   13.6 
5 Colombia                  10   11.4 
6 United States (Hawaii)     8    9.09

Passons à la pratique !

R 中的抽样