Question

我有一个简化的数据集，如下所示：

YEAR = c(2009,2009,2009,2009,2009,2009,2009,2010,2010,2010,2010,2010,2010,2010)
FROM = c("A","C","B","D","B","A","C","A","C","B","A","D","B","A")
TO = c("B","D","C","A","D","C","B","B","A","D","D","C","A","D")
DATA = data.frame(YEAR,FROM,TO)

YEAR    FROM    TO
2009    A       B
2009    C       D
2009    B       C
2009    D       A
2009    B       D
2009    A       C
2009    C       B
2010    A       B
2010    C       A
2010    B       D
2010    A       D
2010    D       C
2010    B       A
2010    A       D

我想要的是两个额外的列，让我们说OCC_FROM和OCC_TO，它是先前行中FROM和TO列中出现的累计计数，按YEAR计算。像这样：

YEAR    FROM    TO  OCC_FROM    OCC_TO
2009    A       B   0           0
2009    C       D   0           0
2009    B       C   1           1
2009    C       A   2           1
2009    B       D   2           1
2009    A       C   2           3
2009    C       B   4           3
2010    A       B   0           0
2010    C       A   0           1
2010    B       D   1           0
2010    A       B   2           2
2010    D       C   1           1
2010    B       A   3           3
2010    A       D   4           2

在Cumulative count in R的帮助下，我设法制作了这个，这显然不是我想要的，因为它没有考虑到YEAR：

DATA$OCC_FROM = sapply(1:length(DATA$FROM),function(i)sum(DATA$FROM[i]==DATA$FROM[1:i]))+sapply(1:length(DATA$FROM),function(i)sum(DATA$FROM[i]==DATA$TO[1:i]))-1
DATA$OCC_TO = sapply(1:length(DATA$TO),function(i)sum(DATA$TO[i]==DATA$FROM[1:i]))+sapply(1:length(DATA$TO),function(i)sum(DATA$TO[i]==DATA$TO[1:i]))-1

YEAR    FROM    TO  OCC_FROM    OCC_TO
2009    A       B   0           0
2009    C       D   0           0
2009    B       C   1           1
2009    C       A   2           1
2009    B       D   2           1
2009    A       C   2           3
2009    C       B   4           3
2010    A       B   3           4
2010    C       A   5           4
2010    B       D   5           2
2010    A       B   5           6
2010    D       C   3           6
2010    B       A   7           6
2010    A       D   7           4

编辑：我也希望能够像以前那样按照YEAR和FROM一起累计累加两列。为简单起见，我将使用OCC_FROM和OCC_TO。像这样：

YEAR    FROM    TO  OCC_FROM    OCC_TO    TOTAL_FROM    TOTAL_TO
2009    A       B   0           0         0             0
2009    C       D   0           0         0             0
2009    B       C   1           1         0             0
2009    C       A   2           1         1             0
2009    B       D   2           1         1             0
2009    A       C   2           3         1             3
2009    C       B   4           3         6             3
2010    A       B   0           0         0             0
2010    C       A   0           1         0             0
2010    B       D   1           0         0             0
2010    A       B   2           2         1             1
2010    D       C   1           1         0             0
2010    B       A   3           3         3             3
2010    A       D   4           2         6             1

Answer 1

你可以尝试

prevCount <- function(x) {
   eq <- outer(x,x,"==")
   eq <- eq & upper.tri(eq)
   eqInt <- ifelse(eq, 1, 0)
   return(apply(eqInt,2,sum))
}
DATA$OCC_FROM <- ave(DATA$FROM, DATA$YEAR, FUN=prevCount )

prevCount是一个函数，它在一年内返回每个元素之前相同的元素数。然后，ave电话会按年适用。

在评论中汇总更正，我们得到

ord <- order(c(1:nrow(DATA), 1:nrow(data)))
targets <- c(data$FROM, data$TO)[ord]
yr <- c(data$YEAR, data$YEAR)[ord]
res <- ave(targets, yr, FUN=prevCount)
data$occ_from <- res[seq(1, length(res), 2)]
data$occ_to <- res[seq(2, length(res), 2)]

此外，prevCount函数可以简化为：

prevCount <- function(x) {ave(x==x, x, FUN=cumsum)}

Answer 2

# Your data - this is not the data at the top of your question but from your 
# solution  [the 'from' and 'to' don't correspond exactly between your question 
# and solution]

YEAR <- c(2009,2009,2009,2009,2009,2009,2009,2010,2010,2010,2010,2010,2010,2010)
FROM <- c("A","C","B","C","B","A","C","A","C","B","A","D","B","A")
TO <- c( "B","D","C","A","D","C","B","B","A","D","B","C","A","D")
mydf <- data.frame(YEAR,FROM,TO)
names(mydf) <- tolower(names(mydf))

#---------------------------------------------------
# Function to get cumulative sum across columns by group
f <- function(from , to){
        # combine the columns 'from' and 'to' alternately
        l <- c(rbind(from , to))

        # Get and sum duplicate values
        dup <- duplicated(l)
        sums <- ave(dup , l , FUN= cumsum)

        # Reshape data & output
        out <- t(matrix(sums ,2))
        colnames(out) <- c("occ_from","occ_to")
        out
      }

# Not considering year 
f(mydf$from , mydf$to)  
# (data.frame(mydf , f(mydf$from , mydf$to) )

# Calculate by year
s <- split(mydf , mydf$year)
d <- do.call(rbind,lapply(s,function(i) f(i[,"from"],i[,"to"])))
(mydf <- data.frame(mydf , d , row.names=NULL))

R - 按因子计算的多列累积计数

2 个答案: