如何为每个分组元素选择随机的非连续日期？

我目前正在尝试为每个分组列选择非连续日期。

换句话说，我有以下数据框：

我基本上想group_by(Site)然后为每个分组站点仅保留 3 个随机非连续日期。例如，如果 HP37P1B 的日期对应于 3 月 12 日、3 月 13 日、3 月 14 日和 3 月 7 日 - 我需要一个仅包含以下内容的数据框（例如）：

HP37P1B 12th March

HP37P1B 14th March

HP37P1B 7th March

到目前为止，我已经尝试了一些使用 stackoverflow 的帖子diff(), ave()，以及lubridate包，但我还没有取得任何成功。

EDIT

根据下面的评论，我试图使这个问题可重现

dput(uniqueSiteDate)

structure(list(Site = c("HP37P1B", "HP37P2B", "HP37P4B", "HP4008U", 
"INME03R", "INME03U", "INOA03R", "IPTO04R", "IPTO04U", "IPTO06R", 
"IPTO06U", "OLCAP2B", "OLCAP3B", "OLCAP5B", "PANMP1B", "PANMP2B", 
"PANMP3B", "STIN02R", "STIN02U", "UPMAP1B", "UPMAP3B", "UPMAP4B", 
"UPMAP5B", "UPMAP6B", "VAR210R", "VAR310R", "VAR310U", "VAR410R", 
"VAR410U", "HP36P1B", "HP36P3B", "HP36P4B", "HP4008R", "INBS04R", 
"INBS04U", "SEL107R", "SEL107U", "SEL207R", "SEL207U", "OLV110R", 
"OLV110U", "OLV208R", "OLV208U", "THEN10U", "HP37P1B", "HP37P2B", 
"HP37P4B", "HP4008U", "INME03R", "INME03U", "INOA03R", "IPTO04R", 
"IPTO04U", "IPTO06R", "IPTO06U", "OLCAP2B", "OLCAP3B", "OLCAP5B", 
"PANMP1B", "PANMP2B", "PANMP3B", "STIN02R", "STIN02U", "UPMAP1B", 
"UPMAP3B", "UPMAP4B", "UPMAP5B", "UPMAP6B", "VAR210R", "VAR310R", 
"VAR310U", "VAR410R", "VAR410U", "OLV110R", "OLV110U", "OLV208R", 
"OLV208U", "THEN10U", "HP37P1B", "HP37P2B", "HP37P4B", "HP4008U", 
"INME03R", "INME03U", "INOA03R", "IPTO04R", "IPTO04U", "IPTO06R", 
"IPTO06U", "OLCAP2B", "OLCAP3B", "OLCAP5B", "PANMP1B", "PANMP2B", 
"PANMP3B", "STIN02R", "STIN02U", "UPMAP1B", "UPMAP3B", "UPMAP4B", 
"UPMAP5B", "UPMAP6B", "VAR210R", "VAR310R", "VAR310U", "VAR410R", 
"VAR410U", "OLV110R", "OLV110U", "OLV208R", "OLV208U", "THEN10U", 
"HP37P1B", "HP37P2B", "HP37P4B", "HP4008U", "INME03R", "INME03U", 
"INOA03R", "IPTO04R", "IPTO04U", "IPTO06R", "IPTO06U", "OLCAP2B", 
"OLCAP3B"), Date = structure(c(18333, 18333, 18333, 18333, 18335, 
18335, 18335, 18338, 18335, 18338, 18335, 18333, 18333, 18333, 
18334, 18334, 18334, 18331, 18331, 18331, 18330, 18330, 18330, 
18330, 18332, 18332, 18332, 18332, 18332, 18325, 18325, 18325, 
18325, 18327, 18327, 18327, 18327, 18327, 18328, 18340, 18340, 
18340, 18340, 18340, 18334, 18334, 18334, 18334, 18336, 18336, 
18336, 18339, 18336, 18340, 18336, 18335, 18334, 18334, 18335, 
18335, 18335, 18332, 18332, 18332, 18331, 18331, 18331, 18331, 
18333, 18333, 18333, 18333, 18333, 18341, 18341, 18341, 18341,
18341, 18335, 18335, 18335, 18335, 18383, 18383, 18383, 18384, 
18384, 18384, 18384, 18385, 18385, 18335, 18342, 18342, 18341, 
18383, 18383, 18345, 18349, 18349, 18349, 18349, 18340, 18339, 
18340, 18341, 18339, 18386, 18386, 18348, 18346, 18347, 18328, 
18328, 18328, 18328, 18390, 18389, 18391, 18392, 18392, 18392, 
18392, 18392, 18392), class = "Date")), row.names = c(NA, -125L
), groups = structure(list(Site = c("HP36P1B", "HP36P3B", "HP36P4B", 
"HP37P1B", "HP37P2B", "HP37P4B", "HP4008R", "HP4008U", "INBS04R", 
"INBS04U", "INME03R", "INME03U", "INOA03R", "IPTO04R", "IPTO04U", 
"IPTO06R", "IPTO06U", "OLCAP2B", "OLCAP3B", "OLCAP5B", "OLV110R", 
"OLV110U", "OLV208R", "OLV208U", "PANMP1B", "PANMP2B", "PANMP3B", 
"SEL107R", "SEL107U", "SEL207R", "SEL207U", "STIN02R", "STIN02U", 
"THEN10U", "UPMAP1B", "UPMAP3B", "UPMAP4B", "UPMAP5B", "UPMAP6B", 
"VAR210R", "VAR310R", "VAR310U", "VAR410R", "VAR410U"), .rows = structure(list(
    30L, 31L, 32L, c(1L, 45L, 79L, 113L), c(2L, 46L, 80L, 114L
    ), c(3L, 47L, 81L, 115L), 33L, c(4L, 48L, 82L, 116L), 34L, 
    35L, c(5L, 49L, 83L, 117L), c(6L, 50L, 84L, 118L), c(7L, 
    51L, 85L, 119L), c(8L, 52L, 86L, 120L), c(9L, 53L, 87L, 121L
    ), c(10L, 54L, 88L, 122L), c(11L, 55L, 89L, 123L), c(12L, 
    56L, 90L, 124L), c(13L, 57L, 91L, 125L), c(14L, 58L, 92L), 
    c(40L, 74L, 108L), c(41L, 75L, 109L), c(42L, 76L, 110L), 
    c(43L, 77L, 111L), c(15L, 59L, 93L), c(16L, 60L, 94L), c(17L, 
    61L, 95L), 36L, 37L, 38L, 39L, c(18L, 62L, 96L), c(19L, 63L, 
    97L), c(44L, 78L, 112L), c(20L, 64L, 98L), c(21L, 65L, 99L
    ), c(22L, 66L, 100L), c(23L, 67L, 101L), c(24L, 68L, 102L
    ), c(25L, 69L, 103L), c(26L, 70L, 104L), c(27L, 71L, 105L
    ), c(28L, 72L, 106L), c(29L, 73L, 107L)), ptype = integer(0), class = c("vctrs_list_of", 
"vctrs_vctr", "list"))), row.names = c(NA, -44L), class = c("tbl_df", 
"tbl", "data.frame"), .drop = TRUE), class = c("grouped_df", 
"tbl_df", "tbl", "data.frame"))

为了回答其他问题，有时每个网站有超过 3 个日期，但有时每个网站只有 1 个日期。但想法是选择给定站点的 n 个非连续日期。换句话说，如果某个特定网站有 4 个日期，我需要 3 个不连续的日期。如果某个特定网站只有 1 个日期，我们就将其保留。

请检查是否达到目的？实际上，根据提供的标准选择最大可能的日期很困难（至少对我来说）。我们可以通过以下策略识别连续和非连续组中的日期。但请考虑一组（例如 3 个连续日期）中的两个场景。如果random样本包含 2 个单元，这些单元可以是连续的，也可以是非连续的。假设如果我们进一步选择奇数 (2) 或偶数 (1) 行，那么在我看来，样本将是判断性的而不是随机的。这是采取的策略——

• 将数据分组
• 通过各组分别进行操作purrr::map_df最后一行绑定数据
• 将数据（现在为组）分为连续日期和非连续日期（每个连续日期将位于其自己的组中）。从每个组中选择唯一的行。
• 最后从每一行中选择三个（或根据小组结果更少）。

library(tidyverse)

df %>% 
  ungroup() %>% 
  group_split(Site) %>% 
  map_df(., ~ .x %>% ungroup() %>%
        arrange(Date) %>%
        mutate(n = 1) %>%
        complete(Date = seq.Date(first(Date), last(Date), by = 'days')) %>%
        group_by(n = cumsum(is.na(n))) %>%
        filter(!is.na(Site)) %>%
        sample_n(1) %>%
        ungroup() %>%
        sample_n(min(n(), 3))) %>%
  select(-n)

# A tibble: 86 x 2
   Date       Site   
   <date>     <chr>  
 1 2020-03-04 HP36P1B
 2 2020-03-04 HP36P3B
 3 2020-03-04 HP36P4B
 4 2020-03-07 HP37P1B
 5 2020-03-12 HP37P1B
 6 2020-03-07 HP37P2B
 7 2020-03-12 HP37P2B
 8 2020-03-07 HP37P4B
 9 2020-03-12 HP37P4B
10 2020-03-04 HP4008R
# ... with 76 more rows

注意：您的 dput 已分组，因此我必须添加ungroup()在代码的第二行，您可以将其删除