How to create lag variables

Question

I want to create lagged variable for a variable pm10 and used the following code. However, I could not get what I wanted. How could I create a lag of pm10?

df2$l1pm10 <- lag(df2$pm10, -1, na.pad = TRUE)
df2$l1pm102 <- lag(df2$pm10, 1)

dput(df2)
structure(list(var1 = 1:10, pm10 = c(26.956073733, NA, 32.838694951, 
39.9560737332, NA, 40.9560737332, 33.956073733, 28.956073733, 
32.348770798, NA), l1pm10 = structure(c(26.956073733, NA, 32.838694951, 
39.9560737332, NA, 40.9560737332, 33.956073733, 28.956073733, 
32.348770798, NA), .Tsp = c(2, 11, 1))), .Names = c("var1", "pm10", 
"l1pm10"), row.names = c("1", "2", "3", "4", "5", "6", "7", "8", 
"9", "10"), class = "data.frame")

Answer 1

You can use head and tail as follows

df2$l1pm10 <- c(tail(df2$pm10, -1), NA)
df2$l1pm102 <- c(NA, head(df2$pm10, -1))
df2
#R>    var1     pm10   l1pm10  l1pm102
#R> 1     1 26.95607       NA       NA
#R> 2     2       NA 32.83869 26.95607
#R> 3     3 32.83869 39.95607       NA
#R> 4     4 39.95607       NA 32.83869
#R> 5     5       NA 40.95607 39.95607
#R> 6     6 40.95607 33.95607       NA
#R> 7     7 33.95607 28.95607 40.95607
#R> 8     8 28.95607 32.34877 33.95607
#R> 9     9 32.34877       NA 28.95607
#R> 10   10       NA       NA 32.34877

# or with transfrom 
transform(df2, l1pm10 = c(tail(pm10, -1), NA), l1pm102 = c(NA, head(pm10, -1)))
#R>    var1     pm10   l1pm10  l1pm102
#R> 1     1 26.95607       NA       NA
#R> 2     2       NA 32.83869 26.95607
#R> 3     3 32.83869 39.95607       NA
#R> 4     4 39.95607       NA 32.83869
#R> 5     5       NA 40.95607 39.95607
#R> 6     6 40.95607 33.95607       NA
#R> 7     7 33.95607 28.95607 40.95607
#R> 8     8 28.95607 32.34877 33.95607
#R> 9     9 32.34877       NA 28.95607
#R> 10   10       NA       NA 32.34877

You can write a general lag function with these two as follows

lag_func <- function(x, k = 1, pad = NA){
  if(k == 0)
    return(x)
  nas <- rep(pad, min(length(x), abs(k)))
  if(k < 0)
    c(tail(x, k), nas) else c(nas, head(x, -k))
}

# use the function to lag the variable
sapply((-11):11, lag_func, x = df2$pm10)
#R>       [,1] [,2] [,3]     [,4]     [,5]     [,6]     [,7]     [,8]     [,9]
#R>  [1,]   NA   NA   NA 32.34877 28.95607 33.95607 40.95607       NA 39.95607
#R>  [2,]   NA   NA   NA       NA 32.34877 28.95607 33.95607 40.95607       NA
#R>  [3,]   NA   NA   NA       NA       NA 32.34877 28.95607 33.95607 40.95607
#R>  [4,]   NA   NA   NA       NA       NA       NA 32.34877 28.95607 33.95607
#R>  [5,]   NA   NA   NA       NA       NA       NA       NA 32.34877 28.95607
#R>  [6,]   NA   NA   NA       NA       NA       NA       NA       NA 32.34877
#R>  [7,]   NA   NA   NA       NA       NA       NA       NA       NA       NA
#R>  [8,]   NA   NA   NA       NA       NA       NA       NA       NA       NA
#R>  [9,]   NA   NA   NA       NA       NA       NA       NA       NA       NA
#R> [10,]   NA   NA   NA       NA       NA       NA       NA       NA       NA
#R>          [,10]    [,11]    [,12]    [,13]    [,14]    [,15]    [,16]    [,17]
#R>  [1,] 32.83869       NA 26.95607       NA       NA       NA       NA       NA
#R>  [2,] 39.95607 32.83869       NA 26.95607       NA       NA       NA       NA
#R>  [3,]       NA 39.95607 32.83869       NA 26.95607       NA       NA       NA
#R>  [4,] 40.95607       NA 39.95607 32.83869       NA 26.95607       NA       NA
#R>  [5,] 33.95607 40.95607       NA 39.95607 32.83869       NA 26.95607       NA
#R>  [6,] 28.95607 33.95607 40.95607       NA 39.95607 32.83869       NA 26.95607
#R>  [7,] 32.34877 28.95607 33.95607 40.95607       NA 39.95607 32.83869       NA
#R>  [8,]       NA 32.34877 28.95607 33.95607 40.95607       NA 39.95607 32.83869
#R>  [9,]       NA       NA 32.34877 28.95607 33.95607 40.95607       NA 39.95607
#R> [10,]       NA       NA       NA 32.34877 28.95607 33.95607 40.95607       NA
#R>          [,18]    [,19]    [,20]    [,21] [,22] [,23]
#R>  [1,]       NA       NA       NA       NA    NA    NA
#R>  [2,]       NA       NA       NA       NA    NA    NA
#R>  [3,]       NA       NA       NA       NA    NA    NA
#R>  [4,]       NA       NA       NA       NA    NA    NA
#R>  [5,]       NA       NA       NA       NA    NA    NA
#R>  [6,]       NA       NA       NA       NA    NA    NA
#R>  [7,] 26.95607       NA       NA       NA    NA    NA
#R>  [8,]       NA 26.95607       NA       NA    NA    NA
#R>  [9,] 32.83869       NA 26.95607       NA    NA    NA
#R> [10,] 39.95607 32.83869       NA 26.95607    NA    NA

Answer 2

I guess a solution for dummies would just be to create a "lagged" version of the vector or column (adding an NA in the first position) and then bind the columns together:

x<-1:10;    #Example vector

x_lagged <- c(NA, x[1:(length(x)-1)]); 

new_x <- cbind(x,x_lagged);

Answer 3

Another alternative is using the shift-function from the data.table package:

library(data.table)
setDT(df2)[, c("l1pm10","l1pm102") := .(shift(pm10, 1L, fill = NA, type = "lag"),
                                        shift(pm10, 1L, fill = NA, type = "lead"))]

this gives:

> df2
    var1     pm10   l1pm10  l1pm102
 1:    1 26.95607       NA       NA
 2:    2       NA 26.95607 32.83869
 3:    3 32.83869       NA 39.95607
 4:    4 39.95607 32.83869       NA
 5:    5       NA 39.95607 40.95607
 6:    6 40.95607       NA 33.95607
 7:    7 33.95607 40.95607 28.95607
 8:    8 28.95607 33.95607 32.34877
 9:    9 32.34877 28.95607       NA
10:   10       NA 32.34877       NA

---

Used data:

df2 <- structure(list(var1 = 1:10, pm10 = c(26.956073733, NA, 32.838694951, 
39.9560737332, NA, 40.9560737332, 33.956073733, 28.956073733, 
32.348770798, NA)), .Names = c("var1", "pm10"), row.names = c("1", 
"2", "3", "4", "5", "6", "7", "8", "9", "10"), class = "data.frame")

Answer 4

I know the question is been accepted but months ago I faced the same problem (in this question) and I wanted to create an homemade lag function. Here is the code:

 df2$lagpm10 <- c(NA, df2$pm10[seq_along(df2$pm10) -1])

 df2
   var1     pm10   l1pm10  lagpm10
1     1 26.95607 26.95607       NA
2     2       NA       NA 26.95607
3     3 32.83869 32.83869       NA
4     4 39.95607 39.95607 32.83869
5     5       NA       NA 39.95607
6     6 40.95607 40.95607       NA
7     7 33.95607 33.95607 40.95607
8     8 28.95607 28.95607 33.95607
9     9 32.34877 32.34877 28.95607
10   10       NA       NA 32.34877

Benchmarks

where Rhertel1 and Rhertel2 are the two lines of code of Rhertel and Sabdem is mine.

Unit: microseconds
     expr     min      lq      mean   median       uq       max neval
 Rhertel1 250.523 257.740 272.07275 260.3355 264.0945  3540.187 10000
 Rhertel2 246.641 253.887 271.77003 256.5380 260.4935 14637.791 10000
   Sabdem  57.762  60.521  65.85315  61.3765  62.6050 12275.979 10000

Answer 5

In base R the function lag() is useful for time series objects. Here you have a dataframe and the situation is somewhat different.

You could try the following, which I admit is not very elegant:

df2$l1pm10 <- sapply(1:nrow(df2), function(x) df2$pm10[x+1])
df2$l1pm102 <- sapply(1:nrow(df2), function(x) df2$pm10[x-1])
#> df2
#   var1     pm10   l1pm10  l1pm102
#1     1 26.95607       NA         
#2     2       NA 32.83869 26.95607
#3     3 32.83869 39.95607       NA
#4     4 39.95607       NA 32.83869
#5     5       NA 40.95607 39.95607
#6     6 40.95607 33.95607       NA
#7     7 33.95607 28.95607 40.95607
#8     8 28.95607 32.34877 33.95607
#9     9 32.34877       NA 28.95607
#10   10       NA       NA 32.34877

An alternative consists in using the Lag() function (with capital "L") from the Hmiscpackage:

library(Hmisc)
df2$l1pm10 <- Lag(df2$pm10, -1)
df2$l1pm102 <- Lag(df2$pm10, +1)
#> df2
#   var1     pm10   l1pm10  l1pm102
#1     1 26.95607       NA       NA
#2     2       NA 32.83869 26.95607
#3     3 32.83869 39.95607       NA
#4     4 39.95607       NA 32.83869
#5     5       NA 40.95607 39.95607
#6     6 40.95607 33.95607       NA
#7     7 33.95607 28.95607 40.95607
#8     8 28.95607 32.34877 33.95607
#9     9 32.34877       NA 28.95607
#10   10       NA       NA 32.34877