在分组数据框中输入缺失值

Question

我在分组数据框中输入缺失值。在 DF 中，Var1 和 Var2 的缺失值是随机的。

数据框按变量分组 Factory:MachineNum。插补是按照这些分组中 Odometer 的顺序完成的。

代码在大约 5-10% 的时间内完美运行。其他 90-95% 的时间它说；

"Error: Column Impute must be length 50 (the group size) or one, not 49".

我觉得可能跟缺失值的随机性有关。也许当至少 1 行共享 2 个缺失值时。

如何使此代码更健壮？

通过多次运行整个代码，您将看到它在大约 5 - 10% 的尝试中有效，最终将生成 Results 数据帧。

library(dplyr)
library(tidyr)

# Create dataframe with some missing values in Var1 and Var2
DF <- data.frame(Factory = c(replicate(150,"Factory_A"), replicate(150,"Factory_B")),
                 MachineNum = c(replicate(100,"Machine01"), replicate(100,"Machine02"), replicate(100,"Machine03")),
                 Odometer = c(replicate(1,sample(1:1000,100,rep=FALSE)), replicate(1,sample(5000:7000,100,rep=FALSE)), replicate(1,sample(10000:11500,100,rep=FALSE))),
                 Var1 =c(replicate(1, sample(c(2:10, NA), 100, rep = TRUE)), replicate(1, sample(c(15:20, NA), 100, rep = TRUE)), replicate(1, sample(c(18:24, NA), 100, rep = TRUE))),
                 Var2 = c(replicate(1, sample(c(110:130, NA), 100, rep = TRUE)), replicate(1, sample(c(160:170, NA), 100, rep = TRUE)), replicate(1, sample(c(220:230, NA), 100, rep = TRUE)))
)




# Variables with missing values that need imputing
cols <- grep('Var', names(DF), value = TRUE)

# Group-wise impution of missing values
library(stinepack)
Models <- DF %>%
  pivot_longer(cols = starts_with('Var')) %>%
  arrange(Factory, MachineNum, name, Odometer) %>%
  group_by(Factory, MachineNum, name) %>%
  mutate(Impute = na.stinterp(value, along = time(Odometer), na.rm = TRUE))



# Convert results from long to wide to visually inspect
Results <- Models %>% 
  group_by(Factory, MachineNum, name) %>% 
  mutate(row = row_number()) %>% 
  tidyr::pivot_wider(names_from = name, values_from = c(value, Impute))

Answer 1

当你在一个组中有前导和尾随 NA 时会发生错误，因为你有 na.rm = TRUE 它会删除它们使组不平衡。

如果您将 na.rm 保持为 FALSE，它将保持 NA 为 NA 和运行而不会出错。

library(dplyr)
library(stinepack)

DF %>%
  pivot_longer(cols = starts_with('Var')) %>%
  arrange(Factory, MachineNum, name, Odometer) %>%
  group_by(Factory, MachineNum, name) %>%
  mutate(Impute = na.stinterp(value, along = time(Odometer), na.rm = FALSE))

在分组数据框中输入缺失值

Imputing missing values in a grouped dataframe

grouping

r

multiple-columns

dplyr

imputation