R 基于相同id的上一行和下一行计算空行的所有数值变量的值

我有一个非常大的data.frame（数千个变量），每行有一个标识符和一年。一行可能出现数年或出现在数据的中间。缺少一年（1997年），我想以某种方式插值所有数值变量的值：

复制上一年（1996年）中存在标识符的所有行 次年（1998年）

对于所有数值变量，计算前一年变量与下一年变量的平均值-对应的特定两行（具有相同标识符）

由于这是一个非常大的数据集，我迫切希望避免使用循环

示例数据：

test_df <- data.frame(id = c(1,2,3,1,3), year = c(96,96,96,98,98), 
                      state = c("MA","MD","NY","MA", "NY"),
                      num1 = c(10,11,22,9,27), num2 = c(11566,32340,97555,14200,100025))
> test_df
  id year state num1   num2
1  1   96    MA   10  11566
2  2   96    MD   11  32340
3  3   96    NY   22  97555
4  1   98    MA    9  14200
5  3   98    NY   27 100025

到目前为止，我所做的是对上一年中与下一年id相同的行进行子设置，并选择数值变量。在计算之后，我将把它们绑定到主数据中

common_ids <- test_df[test_df$year==1996,]
common_ids <- common_ids[test_df[test_df$year==1996,]$id %in% test_df[test_df$year==1998,]$id,]
numeric_vars <- sapply(common_ids,is.numeric)

common_ids[,numeric_vars] <- lapply(common_ids[,numeric_vars], function(x)???)

---

common\u id使用
data.table
和
zoo
您可以从以下内容开始

library(data.table)
library(zoo)

test_df <- data.table(id = c(1,2,3,1,3), year = c(96,96,96,98,98), 
                  state = c("MA","MD","NY","MA", "NY"),
                  num1 = c(10,11,22,9,27), num2 = c(11566,32340,97555,14200,100025))

test_df <- test_df[order(id, year)]

missing.ids <- test_df[, c(NA, id[-.N]), by = id][!is.na(V1),V1]

temp_df <- data.table(id = missing.ids, year = rep(97, length(missing.ids)), state = NA, num1 = NA, num2 = NA)

new.test_df <- rbind(test_df, temp_df)[order(id, year)]

new.test_df[, state := na.locf(state, na.rm = FALSE), by = id]
new.test_df[, `:=` (num1 = na.approx(num1, na.rm = FALSE), num2 = na.approx(num2, na.rm = FALSE)), by = id]

库（data.table）
图书馆（动物园）
test_df您可以在不命名特定变量（有数千个）的情况下执行此操作吗？您可以创建一个列名称向量，并使用
数据在其中循环。table
的
set
。今天没有时间，但也许有人可以更新我的答案
library(data.table)
library(zoo)

test_df <- data.table(id = c(1,2,3,1,3), year = c(96,96,96,98,98), 
                  state = c("MA","MD","NY","MA", "NY"),
                  num1 = c(10,11,22,9,27), num2 = c(11566,32340,97555,14200,100025))

test_df <- test_df[order(id, year)]

missing.ids <- test_df[, c(NA, id[-.N]), by = id][!is.na(V1),V1]

temp_df <- data.table(id = missing.ids, year = rep(97, length(missing.ids)), state = NA, num1 = NA, num2 = NA)

new.test_df <- rbind(test_df, temp_df)[order(id, year)]

new.test_df[, state := na.locf(state, na.rm = FALSE), by = id]
new.test_df[, `:=` (num1 = na.approx(num1, na.rm = FALSE), num2 = na.approx(num2, na.rm = FALSE)), by = id]

library(data.table)
library(zoo)

test_df <- data.table(id = c(1,2,3,1,3), year = c(96,96,96,98,98), 
                  state = c("MA","MD","NY","MA", "NY"),
                  num1 = c(10,11,22,9,27), num2 = c(11566,32340,97555,14200,100025))

test_df <- test_df[order(id, year)]

mynum.cols <- names(test_df)[!(names(test_df) %in% c("id", "year", "state"))]
missing.ids <- test_df[, c(NA, id[-.N]), by = id][!is.na(V1),V1]

temp_df <- data.table(id = missing.ids, year = rep(97, length(missing.ids)), state = NA, 
                  data.table(matrix(NA, nrow = length(missing.ids), ncol = length(mynum.cols), 
                                    dimnames = list(rep(NA, length(missing.ids)), mynum.cols))))

new.test_df <- rbind(test_df, temp_df)[order(id, year)]

new.test_df[, state := na.locf(state, na.rm = FALSE), by = id]

new.test_df[, (mynum.cols) := lapply(.SD, function(x) na.approx(x, na.rm = FALSE)), by = id, .SDcols = mynum.cols]

new.test_df <- new.test_df[order(year, id)]
new.test_df