Obtaining year-on-year percentage change by group

问题

I'm working with a data set corresponding to the extract:

set.seed(1)
df <- data.frame(indicator=runif(n = 100),cohort=letters[1:4],
                 year=rep(1976:2000, each=4))

I would like to generate a variable with percentage year-on-year change for each cohort represented in the data set. I have tried to use the code below (from this discussion):

df$ind_per_chng <- transform(new.col=c(NA,indicator[-1]/indicator[-nrow(df)]-1))

but I'm interested in making it work within each subgroup and generating only one extra column with percentage change instead of set of columns that are presently created:

> head(df)
  indicator cohort year ind_per_chng.indicator ind_per_chng.cohort ind_per_chng.year
1 0.2655087      a 1976              0.2655087                   a              1976
2 0.3721239      b 1976              0.3721239                   b              1976
3 0.5728534      c 1976              0.5728534                   c              1976
4 0.9082078      d 1976              0.9082078                   d              1976
5 0.2016819      a 1977              0.2016819                   a              1977
6 0.8983897      b 1977              0.8983897                   b              1977
  ind_per_chng.new.col
1                   NA
2            0.4015509
3            0.5394157
4            0.5854106
5           -0.7779342
6            3.4544877

---

Edit

To answer the useful comments, the format of the output should correspond to the table below:

There are no other changes to original data.frame with exception of the column that provides value for the percentage change for the selected variable for each cohort across years.

回答1:

I'm not sure I'm correctly understanding what you want the output to look like, but is that what you're after?

library(dplyr)
df2 <- df%>%
    group_by(cohort) %>%
    arrange(year) %>%
    mutate(pct.chg = (indicator - lag(indicator))/lag(indicator))

If you want your percentages on a 0-100 scale instead of 0-1, add 100 * () to that last line, so mutate(pct.chg = 100 * ((indicator - lag(indicator))/lag(indicator))). Here's what the result looks like:

  indicator cohort year    pct.chg
1 0.2655087      a 1976         NA
2 0.2016819      a 1977 -24.039416
3 0.6291140      a 1978 211.933767
4 0.6870228      a 1979   9.204818
5 0.7176185      a 1980   4.453369
6 0.9347052      a 1981  30.250993

来源：https://stackoverflow.com/questions/31812864/obtaining-year-on-year-percentage-change-by-group