R Time series - three month on three month percentage change-CodePudding

I have a time series (df) in R for which I would like to calculate percentage change over varying periods:

month  x 
Jan   1
Feb   4
Mar   5
Apr   3
May   1
Jun   2

I can calculate a month on month percentage change for the series using:

df <- df %>% 
mutate(mom_pct = (count/lag(count)*100-100))

This results in

month  x mom_pct 
Jan   1       NA 
Feb   4      300
Mar   5       25
Apr   3      -40
May   1   -66.67
Jun   2      100

I cannot work out how to produce a three-month on three-month percentage change however (i.e. the sum of the last three months divided by the previous three months). I have tried the following:

df <- df %>% 
mutate("3mo3m_pct" = (rollapplyr(count, 3, sum, fill = NA)/rollapplyr(lag(count, -3), 3, mean, fill = NA))*100-100)

But this returns an error - x n must be a nonnegative integer scalar, not a double vector of length 1.

CodePudding user response：

This was achieved by moving the lag function in the denominator of the 3 month percentage change calculation:

mutate("3mo3m_pct" = (rollapplyr(x, 3, sum, fill = NA)/lag(rollapplyr(x, 3, sum, fill = NA),3)*100-100))

CodePudding user response：

Define a pct function and use it in rollapplyr:

library(dplyr)
library(zoo)

pct <- function(x) 100 * (sum(tail(x, 3)) / sum(head(x, 3)) - 1)
df %>% mutate(pct = rollapplyr(x, 6, pct, fill = NA))

Note

The input in reproducible form

Lines <- "month  x 
Jan   1
Feb   4
Mar   5
Apr   3
May   1
Jun   2"
df <- read.table(text = Lines, header = TRUE, strip.white = TRUE)