Conditional lead variable in R - r

Question

I want to create conditional lead/lag variables that would that capture pre and post years of an agreement signed by countries.
More precisely, I want to create the following variables:
a variable that is =1 in the 4 years pre/before the agreement, 0 otherwise
a variable that is =1 5 years pre the agreement and
a variable that is =1 only after 4 years after the ratification
I have a country-year data (please see below for the sample of the data). The X1 indicates whether a country has signed the agreemen (=1), or not (=0).
The variables I want to create (my expected output) are manually done in the sample data below, labeled as X1_pre4, X1_pre5 and X1_post5. The first captures the 4 years (or up to 4 years) before the agreement is signed. The second captures the 5 years before the agreement is signed. And the last variable captures the 5 years after the agreement is signed (it starts the same year as the agreement is signed, but it's fine if it starts after that, too).
I have been suggested to use some sort of "split-operate-unsplit" construct. But I personally think that this is can be done in dplyr, using the mutate command.
Currently, I've been trying to work with this logic:
data$X1_pre4[data$year<="1972" & data$X1=="0" ] <- "1"
But this is not good enough (far away), as I am not sure how to group by country here. Even if I figure this out, it won't do the work as I have over 100 of X's (agreements). I simply need a code that is much smarter.
data <-
structure(list(country = structure(c(1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L,
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L,
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L,
3L, 3L, 3L, 3L, 3L, 3L), .Label = c("A", "B", "C"), class = "factor"),
year = c(1970L, 1971L, 1972L, 1973L, 1974L, 1975L, 1976L,
1977L, 1978L, 1979L, 1980L, 1981L, 1982L, 1983L, 1984L, 1985L,
1986L, 1987L, 1988L, 1970L, 1971L, 1972L, 1973L, 1974L, 1975L,
1976L, 1977L, 1978L, 1979L, 1980L, 1981L, 1982L, 1983L, 1984L,
1985L, 1986L, 1987L, 1988L, 1970L, 1971L, 1972L, 1973L, 1974L,
1975L, 1976L, 1977L, 1978L, 1979L, 1980L, 1981L, 1982L, 1983L,
1984L, 1985L, 1986L, 1987L, 1988L, 1989L, 1990L, 1991L),
X1 = c(0L, 0L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L), X1_pre4 = c(1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L,
1L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 1L, 1L, 1L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L), X1_pre5 = c(1L, 1L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 1L, 1L, 1L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 1L, 1L,
1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), X1_post4 = c(0L,
0L, 1L, 1L, 1L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 1L, 1L, 1L,
1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 1L, 1L, 1L, 1L, 1L, 0L, 0L, 0L)), class = "data.frame", row.names = c(NA,
-60L))

Answer 1

This would correspond to your logic:
check_pre <- function(idx, k) { pre_vec <- sapply(1:length(idx), function(x) +any(idx[x:(pmin(x + k, length(idx)))] %in% 1)); pre_vec[idx == 1] <- 0; return(pre_vec) }
check_post <- function(idx, k) sapply(1:length(idx), function(x) +any(idx[(pmax(x - k, 1)):x] %in% 1))
df %>%
group_by(country) %>%
mutate(
idx = +( (lag(X1) == 0 & X1 == 1) | row_number() == 1 & X1 == 1),
X1_pre4 = check_pre(idx, 4),
X1_pre5 = check_pre(idx, 5),
X1_post4 = check_post(idx, 4),
idx = NULL
)
Basically we create an index of when the agreement occurred, and then check for the rows before/after this index with custom functions check_pre and check_post.
This is the output:
country year X1 X1_pre4 X1_pre5 X1_post4
1 A 1970 0 1 1 0
2 A 1971 0 1 1 0
3 A 1972 1 0 0 1
4 A 1973 1 0 0 1
5 A 1974 1 0 0 1
6 A 1975 1 0 0 1
7 A 1976 1 0 0 1
8 A 1977 1 0 0 0
9 A 1978 1 0 0 0
10 A 1979 1 0 0 0
11 A 1980 1 0 0 0
12 A 1981 1 0 0 0
13 A 1982 1 0 0 0
14 A 1983 1 0 0 0
15 A 1984 1 0 0 0
16 A 1985 1 0 0 0
17 A 1986 1 0 0 0
18 A 1987 1 0 0 0
19 A 1988 1 0 0 0
20 B 1970 0 0 0 0
21 B 1971 0 0 0 0
22 B 1972 0 0 0 0
23 B 1973 0 0 1 0
24 B 1974 0 1 1 0
25 B 1975 0 1 1 0
26 B 1976 0 1 1 0
27 B 1977 0 1 1 0
28 B 1978 1 0 0 1
29 B 1979 1 0 0 1
30 B 1980 1 0 0 1
31 B 1981 1 0 0 1
32 B 1982 1 0 0 1
33 B 1983 1 0 0 0
34 B 1984 1 0 0 0
35 B 1985 1 0 0 0
36 B 1986 1 0 0 0
37 B 1987 1 0 0 0
38 B 1988 1 0 0 0
39 C 1970 1 0 0 1
40 C 1971 0 0 0 1
41 C 1972 0 0 0 1
42 C 1973 0 0 0 1
43 C 1974 0 0 0 1
44 C 1975 0 0 0 0
45 C 1976 0 0 0 0
46 C 1977 0 0 0 0
47 C 1978 0 0 0 0
48 C 1979 0 0 1 0
49 C 1980 0 1 1 0
50 C 1981 0 1 1 0
51 C 1982 0 1 1 0
52 C 1983 0 1 1 0
53 C 1984 1 0 0 1
54 C 1985 1 0 0 1
55 C 1986 1 0 0 1
56 C 1987 1 0 0 1
57 C 1988 1 0 0 1
58 C 1989 1 0 0 0
59 C 1990 1 0 0 0
60 C 1991 1 0 0 0
It corresponds to your desired output in majority of cases, however from row 39 onwards you don't have it marked as post-agreement - though it occurred in 1970. Either a typo or you'll need to further explain the logic.