Conditional calculation based on other columns lagged values - r

Question

Newbie: I have a dataset where I want to calculate the y-o-y growth of sales of a company. The dataset contains approx. 1000 companies with each different number of years listed on a public stock exchange. The data looks like this:
# gvkey fyear at company name
#22 17436 2010 59393 BASF SE
#23 17436 2011 61175 BASF SE
#24 17436 2012 64327 BASF SE
...
#30 17436 2018 86556 BASF SE
#31 17828 1989 62737 DAIMLER AG
#32 17828 1990 67339 DAIMLER AG
#33 17828 1991 75714 DAIMLER AG
...
#60 17828 2018 281619 DAIMLER AG
I would like to create a new column growth where I calculate the percentage increase of at from e.g. BASF SE (gvkey 17436) from 2010 to 2011, to 2012 and so on. In row #31 the conditional statement is supposed to work that it would not calculate the increase based on values that belong to BASF but rather have a NA value. Therefore the next value in this new column "growth" in row 32 would be the percentage increase of DAIMLER (gvkey 17828) from 62727 to 67339
So far I tried:
if TA$gvkey == lag(TA$gvkey) {mutate(TA, growth = (at - lag(at))/lag(at))} else {NULL}
Basically I tried to condition the calculation on the change of the gvkey identifier as this makes the most sense to me. I believe there is a nicer way of maybe running a loop until the gvkey changes and the continue with the next set of values - but I simply don't know how to code that.
I am very new to R and quite lost. I would appreciate every support! Thank you, guys :)

Answer 1

I do not see a way to do this in one line. Assuming you data is called data you may try:
for(i in data$gvkey){
a = subset(data,data$gvkey==i) # a now contains the data of one company
# calculate pairwise relative difference (assumes sorted years!)
rel_diff = diff(a)/head(a,-1) #diff computes pariwise difference and divide by a ( head(a,-1) removes the last element)
a$growth = c(0,rel_diff) # extend data frame by result, first difference is 0
#output tro somewhere
}
This is a solution with r-base. There might be more efficient ways but this is easy to understand.

Answer 2

In this case the group_by function in dplyr is a good tool to use.By group_by() ing your gv column you will segment out your mutate() call to apply separately for each distinct value of gv. Here is a quick example I made with some dummy data and your same column values:
library(dplyr)
dummyData =
data.frame(gvkey = c(111,111,111,222,222,222),
fyear = c(2010,2012,2011,2010,2011,2013),
at =c(2,4,2,4,5,10)
)
dummyDataTransformed = dummyData %>%
group_by(gvkey) %>%
arrange(fyear) %>% #to make sure we are chronologically in order
mutate(growth = at/lag(at,1) -1) %>% #subtract 1 to get year over year change
ungroup() #I like to ungroup just to make sure i'm not bugging out any calculations I might add further down the line