R- 每组采样随机行，直到达到最大行数

Question

I have a data set from which I want to take a random sample by group up to 30 rows. However, I also want to make sure that at least 1 row for another grouping is included. Additionally, some groups have less than 30 rows, in which case all of the rows for that group should be included. I can't include the exact data set I'm working with because it's proprietary; however, an example for a data frame df would be:

ID|Age|State|Gender|Salary

1 25 CO M 50000
2 34 CO M 72000
3 28 CO M 52000
4 25 CO F 44000
5 25 CA F 55000
6 34 CA F 100000
7 39 CA M 88000
8 34 CA M 59000
... up to 15000 rows

So, I want a random sample of the data set so that no more than 30 rows are given from each state. Then, for each state, I want at least 1 row for each age and gender that exists in the data set. If there are less than 30 age/gender combinations for a given state, but there are more than 30 rows for that state, then the sample should include multiple rows for a given age/gender so that 30 rows are given for that state. If there are less than 30 rows for that state, then I want all the rows in the data set for that state. If there are more than 30 age/gender combinations for a given state, then the sample should have 1 of each up to 30.

Is there a way for me to do this in R?

Answer 1

Here is some code that takes you half the way. First I simulated data, that resembles yours.

df <-
  data.frame(
    ID = 1:1500,
    Age = sample(18:99, 1500, replace = TRUE),
    State = sample(state.abb, 1500, replace = TRUE),
    Gender = sample(c("M", "F"), 1500, replace = TRUE),
    Salary = sample(44:100 * 1000, 1500, replace = TRUE)
  )

Then with group_by() you can create the state grouping, determine the rows per state with mutate() and n() . That information can then be used to draw samples with sample_n() , that adjust to the group size.

library(dplyr)
df %>% 
  group_by(State) %>% 
  mutate(n_state = n()) %>% 
  sample_n(ifelse(n >= 30, 30, n))

This could be extended to calculate further group sizes you mention to use that information to ensure you hit the quotas you are looking for. Unfortunately I do no fully understand what your quotas are from your question.