Cummean 忽略 NA 并将 NA 替换为最后一个有效值

Question

与这篇文章类似计算 cummean() 和 cumsd()，同时忽略 NA 值并填充 NA 但我宁愿坚持

tidyverse

我的数据：

structure(list(season = c("Winter", "Winter", "Winter", "Winter", 
"Winter", "Winter", "Winter", "Winter", "Winter", "Spring", "Spring", 
"Spring", "Spring", "Spring", "Spring", "Spring", "Spring", "Spring"
), tmean = c(NA, 2, 3, 4, NA, NA, NA, 8, NA, 7, 8, 9, NA, NA, 
5, 3, 2, NA)), class = "data.frame", row.names = c(NA, -18L))

season tmean
Winter    NA
Winter     2
Winter     3
Winter     4
Winter    NA
Winter    NA
Winter    NA
Winter     8
Winter    NA
Spring     7
Spring     8
Spring     9
Spring    NA
Spring    NA
Spring     5
Spring     3
Spring     2
Spring    NA

我想要什么：

season tmean cumtmean
Winter    NA       NA
Winter     2        2
Winter     3      2.5
Winter     4        3
Winter    NA        3
Winter    NA        3
Winter    NA        3
Winter     8     4.25
Winter    NA     4.25
Spring     7        7
Spring     8      7.5
Spring     9        8
Spring    NA        8
Spring    NA        8
Spring     5     7.25
Spring     3      6.4
Spring     2     5.66
Spring    NA     5.66

Answer 1

您可以定义自己的辅助函数来考虑 NA 值。

cummean_na <- function(x) {
  num <- cumsum(tidyr::replace_na(x,0))
  denom <- cumsum(!is.na(x))
  if_else(denom!=0, num/denom, NA)
}

然后如果你的data.frame被称为

dd

你可以使用

dplyr

来做

dd %>% 
  mutate(cmtean = cummean_na(tmean), .by=season)

返回预期值

   season tmean   cmtean
1  Winter    NA       NA
2  Winter     2 2.000000
3  Winter     3 2.500000
4  Winter     4 3.000000
5  Winter    NA 3.000000
6  Winter    NA 3.000000
7  Winter    NA 3.000000
8  Winter     8 4.250000
9  Winter    NA 4.250000
10 Spring     7 7.000000
11 Spring     8 7.500000
12 Spring     9 8.000000
13 Spring    NA 8.000000
14 Spring    NA 8.000000
15 Spring     5 7.250000
16 Spring     3 6.400000
17 Spring     2 5.666667
18 Spring    NA 5.666667

Answer 2

你可以这样做：

library(dplyr)
library(tidyr)
df |> 
  group_by(season) |> 
  mutate(cumtmean = replace(tmean, complete.cases(tmean), cummean(na.omit(tmean)))) |> 
  fill(cumtmean)