随机抽样组，得到一个平衡组

Question

这是我的数据框架：

data<- 
 ID     Group    Modules
 1      Male      Physics
 1      Male      Chemistry
 2      Female    Biology
 2      Female    Physics
 2      Female    Chemistry
 3      Male      Physics
 3      Male      Biology
 3      Male      Chemistry
 4      Male      Physics
 4      Male      Biology
 4      Male      Chemistry
 5      Male      Physics
 5      Male      Biology
 5      Male      Chemistry
 6      Male      Physics
 6      Male      Biology
 6      Male      Chemistry
 7      Female    Physics
 7      Female    Biology
 8      Female    Chemistry
 8      Male      Physics
 8      Male      Biology
 9      Male      Chemistry
 9      Male      Physics
 10      Male      Biology
 10      Male      Chemistry
 10      Male      Physics
 11      Male      Biology
 11      Male      Chemistry
 11      Male      Physics
 12      Female    Biology
 12      Female    Chemistry

上述数据中男性(n=9)多于女性(n=3)。我想随机选择3个没有回复的男性，所以我最终会有3个男性和3个女性。我还想保留重复的I，因此我得到的结果将是：

newdata<- 
 ID     Group    Modules
 1      Male      Physics
 1      Male      Chemistry
 2      Female    Biology
 2      Female    Physics
 2      Female    Chemistry
 3      Male      Physics
 3      Male      Biology
 3      Male      Chemistry
 7      Female    Physics
 7      Female    Biology
 7      Female    Chemistry
 12     Female    Physics
 12     Female    Biology
 6      Male      Physics
 6      Male      Biology
 6      Male      Chemistry

下面是我的代码：

samples_per_group<-6

new data<-data%>% group_by(Group)%>%slice(sample(n(),min(sampples_per_group, n())))%>%ungroup()

当我尝试运行它时，它选择6个样本大小(每组3个)，但是它只从每个参与者中获取一行，而不是返回该参与者的所有行。基本上，我希望在每个组上选择3个id，而不考虑该id重复的次数。欢迎任何帮助。谢谢

Answer 1

如果要对ID进行采样，则需要获取ID并对其进行采样：

groups = data %>%
  distinct(ID, Group) %>%
  group_by(Group) %>%
  summarize(group_size = n())

smallest_group = min(groups$group_size)

groups %>%
  group_by(Group) %>%
  sample_n(size = smallest_group) %>%
  ungroup() %>%
  left_join(data)

像上面这样的东西应该可以工作。在dplyr链中跨组获取单个数字是困难的--这是可行的--但我认为更清晰的方法是打破管道并提取数字。我们按组采样3个(或任意多个)ID，然后连接回主数据，以获得与这些ID对应的所有行。