Question

我在data.frame中有一个不规则间隔的时间序列。如何在每个事件中获得一行，以获得每个事件的最大值？（不仅仅是每个事件的最大值。）

如果事件间隔超过一定时间（例如三天），则事件被定义为不同。以下是一些虚假数据：

set.seed(42)
x <- data.frame(date=as.Date("2017-08-01")+cumsum(ceiling(rexp(200, rate=0.2))),
                value=round(cumsum(rnorm(200, sd=8)))+500)
plot(x, type="o", pch=16, cex=0.6, las=1)
head(x, 20)

Answer 1

# Time differences between observations:
x$diff <- c(0, as.numeric(diff(x$date)) )
# distinct event if more than 3 days apart:
x$event <- cumsum(x$diff>3) 
# simply get maximum value per event:
tapply(x$value, x$event, max, na.rm=TRUE)

# Get one observation row per event (the maximum):
x$max <- unlist(tapply(x$value, x$event, FUN=function(v){
  out <- rep(0, length(v))
  out[which.max(v)] <- 1 # select first maximum value if there are ties
  out
  }))

head(x, 20) 

# independent event maxima rows:
x[x$max==1, 1:2]

可以像这样获得每小时的时差：

diffs <- as.difftime(diff(x$date))
units(diffs) <- "hours"
diffs <- as.numeric(diffs)

事件检测：每个事件获取行

1 个答案: