Gestratificeerde en gewogen aselecte steekproeven

Steekproeven in R

Richie Cotton

Data Evangelist at DataCamp

Koffies per land

Koffiebonen gegroepeerd per herkomstland.

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 De dataset noemt Hawaï en Taiwan gemakshalve landen, omdat het belangrijke koffieregio’s zijn.

Filteren op 6 landen

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)

Of, gelijkwaardig

coffee_ratings_top <- coffee_ratings %>%
  semi_join(top_counts)

¹ Leer over semi-joins in “Joining Data with dplyr”, hoofdstuk 3.

Aantallen van een eenvoudige aselecte steekproef

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

Aantallen vergelijken

Populatie

# 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

Steekproef van 10%

# 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

Proportioneel gestratificeerde steekproef

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

Gelijke aantallen per stratum

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

Gewogen aselecte steekproef

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))

Gewogen steekproef van 10%

# 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

Laten we oefenen!

Steekproeven in R