我想找到两组之间变量的最接近匹配(最小差异),但如果已经进行了最接近的匹配,请继续下一个最接近的匹配,直到进行了 n 个匹配。
我使用这个答案(下图)中的代码来查找所有组的每个成对分组的Samples
之间最接近的value
匹配(即Location
由VAR
)。
但是,有很多重复,Sample.x
1、2 和 3 的最佳匹配项可能都是Sample.y
1。
相反,我想找到Sample.x
2、3 的下一个最接近的匹配项,直到我指定了不同(Sample.x
-Sample.y
)匹配的数量。但是Sample.x
的顺序并不重要,我只是在寻找给定分组的Sample.x
和Sample.y
之间的前n个匹配项。
我尝试使用如下所示dplyr::distinct
执行此操作。但是我不确定如何使用不同的Sample.y
条目来过滤数据帧,然后再次按最小DIFF
过滤数据帧。但是,这不一定会导致唯一的Sample
配对。
有没有一种聪明的方法可以在 R 中使用 dplyr 实现这一点?这种类型的操作有名称吗?
df01 <- data.frame(Location = rep(c("A", "C"), each =10),
Sample = rep(c(1:10), times =2),
Var1 = signif(runif(20, 55, 58), digits=4),
Var2 = rep(c(1:10), times =2))
df001 <- data.frame(Location = rep(c("B"), each =10),
Sample = rep(c(1:10), times =1),
Var1 = c(1.2, 1.3, 1.4, 1.6, 56, 110.1, 111.6, 111.7, 111.8, 120.5),
Var2 = c(1.5, 10.1, 10.2, 11.7, 12.5, 13.6, 14.4, 18.1, 20.9, 21.3))
df <- rbind(df01, df001)
dfl <- df %>% gather(VAR, value, 3:4)
df.result <- df %>%
# get the unique elements of Location
distinct(Location) %>%
# pull the column as a vector
pull %>%
# it is factor, so convert it to character
as.character %>%
# get the pairwise combinations in a list
combn(m = 2, simplify = FALSE) %>%
# loop through the list with map and do the full_join
# with the long format data dfl
map(~ full_join(dfl %>%
filter(Location == first(.x)),
dfl %>%
filter(Location == last(.x)), by = "VAR") %>%
# create a column of absolute difference
mutate(DIFF = abs(value.x - value.y)) %>%
# grouped by VAR, Sample.x
group_by(VAR, Sample.x) %>%
# apply the top_n with wt as DIFF
# here I choose 5,
# and then hope that this is enough to get a smaller n of final matches
top_n(-5, DIFF) %>%
mutate(GG = paste(Location.x, Location.y, sep="-")))
res1 <- rbindlist(df.result)
res2 <- res1 %>% group_by(GG, VAR) %>% distinct(Sample.y)
res3 <- res2 %>% group_by(GG, VAR) %>% top_n(-2, DIFF)
我通过删除行top_n(-5, DIFF) %>%
来编辑上面产生df.result
的代码。现在res1
包含Sample.x
和Sample.y
的所有匹配项。
然后我在下面的代码中使用了res1
。这可能并不完美,但它所做的是为Sample.x
的第一个条目找到最接近Sample.y
匹配项。然后,从数据帧中筛选出这两个Samples
。匹配将重复,直到找到Sample.y
的每个唯一值的匹配项。结果可能会有所不同,具体取决于首先进行的匹配。
fun <- function(df) {
HowMany <- length(unique(df$Sample.y))
i <- 1
MyList_FF <- list()
df_f <- df
while (i <= HowMany){
res1 <- df_f %>%
group_by(grp, VAR, Sample.x) %>%
filter(DIFF == min(DIFF)) %>%
ungroup() %>%
mutate(Rank1 = dense_rank(DIFF))
res2 <- res1 %>% group_by(grp, VAR) %>% filter(rank(Rank1, ties.method="first")==1)
SY <- as.numeric(res2$Sample.y)
SX <- as.numeric(res2$Sample.x)
res3 <- df_f %>% filter(Sample.y != SY) # filter Sample.y
res4 <- res3 %>% filter(Sample.x != SX) # filter Sample.x
df_f <- res4
MyList_FF[[i]] <- res2
i <- i + 1
}
do.call("rbind", MyList_FF) # https://stackoverflow.com/a/55542822/1670053
}
df <- res1
MyResult <- df %>%
dplyr::group_split(grp, VAR) %>%
map_df(fun)