我遇到了以下问题,但无济于事:
d <- data.frame(value = 1:4, row.names = c("abc", "abcd", "ef", "gh"))
value
abc 1
abcd 2
ef 3
gh 4
l <- nrow(d)
wordmat <- matrix(rep(NA, l^2), l, l, dimnames = list(row.names(d), row.names(d)))
for (i in 1:ncol(wordmat)) {
rid <- agrep(colnames(wordmat)[i], rownames(wordmat), max = 0)
d$matchid[i] <- paste(rid, collapse = ";")
}
# desired output:
(d_agg <- data.frame(value = c(3, 3, 4), row.names = c("abc;abcd", "ef", "gh")))
value
abc;abcd 3
ef 3
gh 4
有这个功能吗?
答案 0 :(得分:1)
这适用于您的示例,但可能需要调整真实的东西:
d <- data.frame(value = 1:4, row.names = c("abc", "abcd", "ef", "gh"))
rowclust <- hclust(as.dist(adist(rownames(d))), method="single")
rowgroups <- cutree(rowclust, h=1.5)
rowagg <- aggregate(d, list(rowgroups), sum)
rowname <- unclass(by(rownames(d), rowgroups, paste, collapse=";"))
rownames(rowagg) <- rowname
rowagg
Group.1 value
abc;abcd 1 3
ef 2 3
gh 3 4
答案 1 :(得分:1)
这是一个可能的解决方案,您可以根据自己的需要进行修改。
一些注意事项:
rownames()
,尤其是在最后阶段,所以这取决于您是否乐意将行名复制为新变量。这是功能。
matches <- function(data, ...) {
temp = vector("list", nrow(data))
for (i in 1:nrow(data)) {
temp1 = agrep(data$RowNames[i], data$RowNames, value = TRUE, ...)
temp[[i]] = data.frame(RowNames = paste(temp1, collapse = "; "),
value = sum(data[temp1, "value"]))
}
temp = do.call(rbind, temp)
temp[!duplicated(temp$RowNames), ]
}
请注意,该函数需要一个名为RowNames
的列,因此我们将创建该列,然后测试该函数。
d <- data.frame(value = 1:4, row.names = c("abc", "abcd", "ef", "gh"))
d$RowNames <- rownames(d)
matches(d)
# RowNames value
# 1 abc; abcd 3
# 3 ef 3
# 4 gh 4
matches(d, max.distance = 2)
# RowNames value
# 1 abc; abcd 3
# 3 abc; abcd; ef; gh 10
matches(d, max.distance = 4)
# RowNames value
# 1 abc; abcd; ef; gh 10