不规则时间序列的 dplyr 自定义滞后函数 [英] dplyr custom lag function for irregular time series

问题描述

我有一个不规则的时间序列，其中数据集中存在空白.此外，数据被分组.滞后函数我已经能够通过观察找到滞后(所以他们找到了数据集中的先前记录)，但我想指定一个时间变量并通过匹配滞后时间来计算滞后.这个问题:R 滞后/领先不规则时间序列数据类似的事情.但是，我无法使用 zoo 解决方案(我有某种包不兼容，根本无法使用 zoo)并且未能成功制作 data.table 解决方案成为足够灵活的解决方案，可以用作具有滞后量作为输入和分组数据容量的函数.

I have an irregular time series, where there are gaps in the dataset. Further, the data is grouped. The lag functions I have been able to find lag by observation (so they find the prior record in the dataset), but I want to specify a time variable and have the lag calculated by matching the lagged time. This question: R lag/lead irregular time series data is doing a similar thing. However, I can't use zoo solution (I have some sort of package incompatibility and can't use zoo at all) and have been unsuccessful in making the data.table solution into something sufficiently flexible to use as a function with lag amount as an input and the capacity for grouped data.

测试数据:

testdf <- data.frame(group = c(1,1,1,1,1,2,2,2,2,2),
                 counter = c(1,2,3,5,6,7,8,9,11,12),
                 xval = seq(100, 1000, 100))
lagamount <- 1

输出应该是向量:NA 100 200 NA 400 NA 600 700 NA 900

The output should be the vector: NA 100 200 NA 400 NA 600 700 NA 900

这是我目前使用的:

library(dplyr)
testout <- group_by(testdf, group) %>%
  mutate(testout = function(x) which((testdf$counter - x) == lagamount))

这给了我一个数据类型错误，即某些(未指定的)不是向量.

This gives me a datatype error that something (unspecified) is not a vector.

有没有办法让这个构建工作?或者，我怎么会滞后于具有分组变量的不规则时间序列?

Is there a way to make this construction work? Alternatively, how could I lag with irregular time series with grouped variables?

推荐答案

在 dplyr 中执行此操作的唯一方法，无需使用 do，将首先明确隐含的缺失值，然后过滤掉它们.

The only way to do this within dplyr, whithout resorting to using do, would be to first make implicit missing values explicit, and filter them out afterwards.

提供一个要变异的向量，然后使用 ifelse(或者可能是新的 dplyr::if_else)来检查滞后是否是您想要的.示例:

Supply a vector to mutate, and use ifelse (or perhaps the new dplyr::if_else) to check whether the lag is what you want it to be. Example:

library(tidyr)
lagamount <- 2

testout <- group_by(testdf, group) %>%
  complete(group, counter = min(counter):max(counter)) %>% 
  mutate(testout = if_else(counter - lag(counter, lagamount) == lagamount, 
                           lag(xval, lagamount), 
                           NA_real_)) %>% 
  filter(!is.na(xval))

产生:

Source: local data frame [10 x 4]
Groups: group [2]

   group counter  xval testout
   <dbl>   <dbl> <dbl>   <dbl>
1      1       1   100      NA
2      1       2   200      NA
3      1       3   300     100
4      1       5   400     300
5      1       6   500      NA
6      2       7   600      NA
7      2       8   700      NA
8      2       9   800     600
9      2      11   900     800
10     2      12  1000      NA

这篇关于不规则时间序列的 dplyr 自定义滞后函数的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！