r - 最后一次值变化

Question

我有一个 xts 对象:

df <- structure(c(0L, 0L, 1L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 1L, 1L, 1L, 
  0L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L),
  .Dim = c(10L, 3L), .Dimnames = list(NULL, NULL),
  index = structure(c(790387200, 790473600, 790560000, 790819200, 790905600,
  790992000, 791078400, 791164800, 791424000, 791510400), tzone = "UTC",
  tclass = "Date"), .indexCLASS = "Date", tclass = "Date", .indexTZ = "UTC",
  tzone = "UTC", class = c("xts", "zoo"))
df
#            [,1] [,2] [,3]
# 1995-01-18    0    1    1
# 1995-01-19    0    1    1
# 1995-01-20    1    1    1
# 1995-01-23    1    0    1
# 1995-01-24    1    1    1
# 1995-01-25    0    1    1
# 1995-01-26    0    1    0
# 1995-01-27    0    1    1
# 1995-01-30    0    1    1
# 1995-01-31    0    0    1

设 1 等于 TRUE，0 等于 FALSE。虽然这只是数据的一小部分，但我想找到 0 变为 1 时的最新(最后)事件。所以对于第一列，这发生在 1995-1-20，第二列发生在 1995-01-24 , 和 1995-01-27 中的第三列。

我试过了

max.col(t(df),"last")

但这会返回最近出现的 1。

实现此目标的最佳方法是什么？

Answer 1

您可以扩展您的max.col 想法以包含diff:

max.col(t(sapply(df[,-1], diff)), "last") + 1

上面假设一个第一列是日期的data.frame。对于 xts 对象(行名称中包含日期)，执行:

max.col(t(diff(df)[-1]), "last") + 1

编辑 更正@G.Grothendieck 指出的问题:

df.diff = t(diff(df)[-1])
max.col(df.diff, "last") + 1 + (rowSums(df.diff > 0) == 0)
# or put an ifelse instead and assign NA or 0 or whatever you like