Question

我想测试不同长度的移动平均线和因变量之间的相关性。我已经编写了一个for循环来完成工作，但很明显，循环不是理想的解决方案。我想知道是否有人可以给我一些关于如何替换这个for循环的功能的指针，并应用为更优雅的解决方案？我提供了代码和测试数据。

library(zoo)

# a function that calculates the correlation between moving averages for 
different lengths of window
# the input functions are "independent": the variable over which to apply the 
moving function
# "dependent": the output column, "startLength": the shortest window length, 
"endLength" the longest window length
# "functionType": the function to apply (mean, sd, etc.)

MovingAverageCorrelation <- function(indepedent, depedent, startLength, endLength, functionType) {
# declare an matrix for the different rolling functions and a correlation vector
avgMat <- matrix(nrow = length(depedent), ncol = (endLength-startLength+1)) 
corVector <- rep(NA, ncol(avgMat))
# run the rollapply function over the data and calculate the corresponding correlations
for (i in startLength:endLength) {
   avgMat[, i] <- rollapply(indepedent, width = i, FUN = functionType, 
                         na.rm = T, fill = NA, align = "right")
   corVector[i] <- cor(avgMat[, i], depedent, use = "complete.obs")
  }
return(corVector)
}

# set test data

set.seed(100)
indVector <- runif(1000)
depVector <- runif(1000)

# run the function over the data

cor <- MovingAverageCorrelation(indVector, depVector, 1, 100, "mean")

谢谢！

Answer 1

尝试sapply：

sapply(1:100, function(i) cor(rollapplyr(indVector, i, mean, na.rm = TRUE, fill = NA), 
        depVector, use = "complete.obs"))

如果您的输入中没有NA，这将起作用并且速度要快得多：

sapply(1:100, function(i) cor(rollmeanr(indVector, i, fill = NA), depVector, use = "comp"))

用R中的apply替换循环的滚动平均值

1 个答案: