Question

很难复制，但可以说：

我有一个数据框，其中有107列关于气象站的每月风速（1961年以来的每月数据）。我想针对时间序列中的Breakpoins标准化每一列的数据。例如，如果某列具有1971-04年的第一个BP，则应使用从第一个记录（1961-01）到第一个BP（1971-04）的均值和标准差进行标准化。如果第二个BP是在1989-05年，则平均值和sd必须从第一个BP到第二个BP。然后，我用新获得的数据替换原始数据。

我执行的代码如下：

library(strucchange)

df <- data.frame(date = seq(as.Date('1961-01-01'),length.out = 700, by = 'months' ), A = rnorm(700, 0, 8.5), 
                 B = rnorm(700, 0, 9.5), C = rnorm(700, 0, 12.4), D = rnorm(700, 0, 5.5)) # create a time series

df[c(2,3,4)][340:560,] <- df[c(2,3,4)][340:560,] + rnorm(12, 87.4, 121.4) # insert some breakpoints for the first 4 columns

bp <- breakpoints(df[,5] ~ 1)
bp <- bp$breakpoints                   

for (a in names(df[,2:ncol(df)])){
  print(a)
  stat <- df[,c('date',a)]
  bp <- breakpoints(stat[,2] ~ 1)
  bp <- bp$breakpoints  
  dates <- stat[bp,] # create a df with the breakpoints
  if(nrow(dates==0)){ # condition if a column does not have any BP
    stat[,2] <- (stat[,2] - mean(stat[,2], na.rm = T))/sd(stat[,2], na.rm = T)
    df[,a] <- stat[,2]
  } else { #if there are BP in the data ...
    for (b in 1:nrow(dates)){
      print(b)
      if(b==1){ #calculate the mean and sd from the first row
        substr <- stat[stat$date >= min(stat$date) & stat$date < dates$date[b],]
        substr[,2] <- (substr[,2] - mean(substr[,2], na.rm = T))/sd(substr[,2], na.rm = T)
        df[,a][df$date >= min(df$date) & df$date < dates$date[b]] <- substr[,2]
      } else if (b == nrow(dates)){ #calculate the mean and sd till the last
        substr <- stat[stat$date >= dates$date[b-1] & stat$date <= max(stat$date),]
        substr[,2] <- (substr[,2] - mean(substr[,2], na.rm = T))/sd(substr[,2], na.rm = T)
        df[,a][df$date >= dates$date[b-1] & df$date < max(stat$date)] <- substr[,2]
      } else if (b > 1) { # if the BP are neither the first or the last one
        substr <- stat[stat$date >= dates$date[b-1] & stat$date < dates$date[b],]
        substr[,2] <- (substr[,2] - mean(substr[,2], na.rm = T))/sd(substr[,2], na.rm = T)
        df[,a][df$date >= dates$date[b-1] & df$date < dates$date[b]] <- substr[,2]
      }
    }
  }
}

但是，当我手动进行验证时，这些值是错误的。有没有人有任何技巧来简化此代码？（并使其正常工作）？谢谢

通过使用断点获得的子集标准化数据框中的列

0 个答案: