How can I automatically create n lags in a timeseries?

Question

I have a dataframe with a column t. I want to create n lagged columns that has names like t-1,t-2 etc..

  year      t  t-1 t-2
19620101    1   NA  NA
19630102    2   1   NA
19640103    3   2   1
19650104    4   3   2
19650104    5   4   3
19650104    6   5   4

My idea is that I will do it in four steps:

• A loop for the column names using "paste"
• A loop for the temporary dataframes for lagged columns using "paste"
• A loop for creating the lagged columns
• cbind them.

But I am not able to proceed with the code. Something rough:

df_final<-lagged(df="odd",n=3)

lagged<-function(df,n){
   df<-zoo(df)
   lags<-paste("A", 1:n, sep ="_")
   for (i in 1:5) {
     odd<-as.data.frame(lag(odd$OBS_Q,-1*i,na.pad =  TRUE))

   #Cbind here
   } 

I am stuck in writing this function. Could you please show some way? Or a different simpler way of doing this....

Reference: Basic lag in R vector/dataframe

---

Addendum:

Real data:

x<-structure(list(DATE = 19630101:19630104, PRECIP = c(0, 0, 0,0), 
               OBS_Q = c(1.61, 1.48, 1.4, 1.33), swb = c(1.75, 1.73, 1.7,1.67), 
               gr4j = c(1.9, 1.77, 1.67, 1.58), isba = c(0.83, 0.83,0.83, 0.83), 
               noah = c(1.31, 1.19, 1.24, 1.31), sac = c(1.99,1.8, 1.66, 1.57), 
               swap = c(1.1, 1.05, 1.08, 0.99), vic.mm.day. = c(2.1,1.75, 1.55, 1.43)), 
          .Names = c("DATE", "PRECIP", "OBS_Q", "swb","gr4j", "isba", "noah", "sac", "swap", "vic.mm.day."), 
          class = c("data.table","data.frame"), row.names = c(NA, -4L))

The column to be lagged is OBS_Q.

Answer 1

I might build something around base R's embed()

x <- c(rep(NA,2),1:6)
embed(x,3)
#      [,1] [,2] [,3]
# [1,]    1   NA   NA
# [2,]    2    1   NA
# [3,]    3    2    1
# [4,]    4    3    2
# [5,]    5    4    3
# [6,]    6    5    4

Perhaps something like this:

f <- function(x, dimension, pad) {
    if(!missing(pad)) {
        x <- c(rep(pad, dimension-1), x)
    }
    embed(x, dimension)
}
f(1:6, dimension=3, pad=NA)
#      [,1] [,2] [,3]
# [1,]    1   NA   NA
# [2,]    2    1   NA
# [3,]    3    2    1
# [4,]    4    3    2
# [5,]    5    4    3
# [6,]    6    5    4

Answer 2

If you are looking for efficiency, try data.tables new shift function

library(data.table) # V >= 1.9.5
n <- 2
setDT(df)[, paste("t", 1:n) := shift(t, 1:n)][]
#    t t 1 t 2
# 1: 1  NA  NA
# 2: 2   1  NA
# 3: 3   2   1
# 4: 4   3   2
# 5: 5   4   3
# 6: 6   5   4 

Here you can set any name for your new columns (within paste) and you also don't need to bind this back to the original as this updates your data set by reference using the := operator.

Answer 3

1) lag.zoo The lag.zoo function in the zoo package can accept a vector of lags. Here we want the 0th lag, the -1 lag and the -2 lag:

library(zoo)
cbind(DF[-2], coredata(lag(zoo(DF$t), 0:-2)))

giving:

      year lag0 lag-1 lag-2
1 19620101    1    NA    NA
2 19630102    2     1    NA
3 19640103    3     2     1
4 19650104    4     3     2
5 19650104    5     4     3
6 19650104    6     5     4

which is as you have in the question but are you sure that that is what you want? The last three rows all have the same date so the 4th row, for example, is being lagged to the same date.

2) head Defining a simple Lag function we can do this using only the base of R:

Lag <- function(x, n = 1) c(rep(NA, n), head(x, -n))  # n > 0

data.frame(DF, `t-1` = Lag(DF$t), `t-2` = Lag(DF$t, 2), check.names = FALSE)

giving:

      year t t-1 t-2
1 19620101 1  NA  NA
2 19630102 2   1  NA
3 19640103 3   2   1
4 19650104 4   3   2
5 19650104 5   4   3
6 19650104 6   5   4

Note: We used this as our data frame:

DF <- data.frame(year = c(19620101, 19630102, 19640103, 19650104, 19650104,
                        19650104), t = 1:6)