r - calculate using next non-na value in data frame column

Question

I have some data in a dataframe, and I would like to calculate the percentage change between the month value. The problem is I have NA in some entries and it throws of the calculation.

       irm     code        price    pct.change
1  201807 511130F075A04      4.6600   2.192982
2  201806 511130F075A04      4.5600   1.333333
3  201805 511130F075A04      4.5000 -13.461538
4  201804 511130F075A04      5.2000         NA
5  201803 511130F075A04          NA         NA
6  201802 511130F075A04      4.9100   1.867220
7  201801 511130F075A04      4.8200  -5.304519
8  201712 511130F075A04      5.0900   2.414487
9  201711 511130F075A04      4.9700  -3.307393
10 201710 511130F075A04      5.1400         NA
11 201709 511130F075A04          NA         NA
12 201708 511130F075A04      5.2900   2.918288
13 201707 511130F075A04      5.1400  66.553255
14 201706 511130F075A04      3.0861 -10.664351
15 201705 511130F075A04      3.4545  -7.241824

The problem is in row 4 and row 10 in the pct.change column. They are NA but I would like them to be calculated using the latest value of price that is not NA . The desired output would be (see rows 4 and 10):

       irm     code        price    pct.change
1  201807 511130F075A04      4.6600   2.192982
2  201806 511130F075A04      4.5600   1.333333
3  201805 511130F075A04      4.5000 -13.461538
**4  201804 511130F075A04      5.2000   5.906314**
5  201803 511130F075A04          NA         NA
6  201802 511130F075A04      4.9100   1.867220
7  201801 511130F075A04      4.8200  -5.304519
8  201712 511130F075A04      5.0900   2.414487
9  201711 511130F075A04      4.9700  -3.307393
**10 201710 511130F075A04      5.1400  -2.835539**
11 201709 511130F075A04          NA         NA
12 201708 511130F075A04      5.2900   2.918288
13 201707 511130F075A04      5.1400  66.553255
14 201706 511130F075A04      3.0861 -10.664351
15 201705 511130F075A04      3.4545  -7.241824

I had tried the standard (x/lead(x) - 1)*100 and several variations using (x/lag(which(!is.na(lead(x)) but I seem to be missing something. Is there a straightforward way to do it in base or even dplyr ? I don't want to replace the NAs, I want to keep them.

Answer 1

@LAP's comment is probably the best way to do it. The syntax is a little better with data.table

library(data.table)
setDT(df)

df[!is.na(price), pct.change := 100*(price/shift(price, type = 'lead') - 1)]

#        irm          code  price pct.change
#  1: 201807 511130F075A04 4.6600   2.192982
#  2: 201806 511130F075A04 4.5600   1.333333
#  3: 201805 511130F075A04 4.5000 -13.461538
#  4: 201804 511130F075A04 5.2000   5.906314
#  5: 201803 511130F075A04     NA         NA
#  6: 201802 511130F075A04 4.9100   1.867220
#  7: 201801 511130F075A04 4.8200  -5.304519
#  8: 201712 511130F075A04 5.0900   2.414487
#  9: 201711 511130F075A04 4.9700  -3.307393
# 10: 201710 511130F075A04 5.1400  -2.835539
# 11: 201709 511130F075A04     NA         NA
# 12: 201708 511130F075A04 5.2900   2.918288
# 13: 201707 511130F075A04 5.1400  66.553255
# 14: 201706 511130F075A04 3.0861 -10.664351
# 15: 201705 511130F075A04 3.4545         NA

Answer 2

in Base R you can decide to replace:

 a = which(is.na(df$price))-1
 transform(df,pct.change=replace(pct.change,a,100*(price[a]/price[a+2]-1)))
      irm          code  price pct.change
1  201807 511130F075A04 4.6600   2.192982
2  201806 511130F075A04 4.5600   1.333333
3  201805 511130F075A04 4.5000 -13.461538
4  201804 511130F075A04 5.2000   5.906314
5  201803 511130F075A04     NA         NA
6  201802 511130F075A04 4.9100   1.867220
7  201801 511130F075A04 4.8200  -5.304519
8  201712 511130F075A04 5.0900   2.414487
9  201711 511130F075A04 4.9700  -3.307393
10 201710 511130F075A04 5.1400  -2.835539
11 201709 511130F075A04     NA         NA
12 201708 511130F075A04 5.2900   2.918288
13 201707 511130F075A04 5.1400  66.553255
14 201706 511130F075A04 3.0861 -10.664351
15 201705 511130F075A04 3.4545  -7.241824