creating index conditioned on value in other column; differences over time - r

Question

I am struggling with the following problem:
The dataframe below contains the development of a value over time for various ids. What i try to get is the increase/decrease of these values based on a the value in a year when event occurred. Several events can occur within one id, so a new event becomes the new baseline year for the id.
To make things clearer, I also add the outcome I want below
What i have
id value year event
a 100 1950 NA
a 101 1951 NA
a 102 1952 NA
a 103 1953 NA
a 104 1954 NA
a 105 1955 X
a 106 1956 NA
a 107 1957 NA
a 108 1958 NA
a 107 1959 Y
a 106 1960 NA
a 105 1961 NA
a 104.8 1962 NA
a 104.2 1963 NA
b 70 1970 NA
b 75 1971 NA
b 80 1972 NA
b 85 1973 NA
b 90 1974 NA
b 60 1975 Z
b 59 1976 NA
b 58 1977 NA
b 57 1978 NA
b 56 1979 NA
b 55 1980 W
b 54 1981 NA
b 53 1982 NA
b 52 1983 NA
b 51 1984 NA
What I am looking for
id value year event index growth
a 100 1950 NA 0
a 101 1951 NA 0
a 102 1952 NA 0
a 103 1953 NA 0
a 104 1954 NA 0
a 105 1955 X 1 1
a 106 1956 NA 2 1.00952381
a 107 1957 NA 3 1.019047619
a 108 1958 NA 4 1.028571429
a 107 1959 Y 1 1 #new baseline year
a 106 1960 NA 2 0.990654206
a 105 1961 NA 3 0.981308411
a 104.8 1962 NA 4 0.979439252
a 104.2 1963 NA 5 0.973831776
b 70 1970 NA 6
b 75 1971 NA 7
b 80 1972 NA 8
b 85 1973 NA 9
b 90 1974 NA 10
b 60 1975 Z 1 1
b 59 1976 NA 2 0.983333333
b 58 1977 NA 3 0.966666667
b 57 1978 NA 4 0.95
b 56 1979 NA 5 0.933333333
b 55 1980 W 1 1 #new baseline year
b 54 1981 NA 2 0.981818182
b 53 1982 NA 3 0.963636364
b 52 1983 NA 4 0.945454545
b 51 1984 NA 5 0.927272727
What I tried
This and this post were quite helpful and I managed to create differences between the years, however, I fail to reset the base year (index) when there is a new event. Furthermore, I am doubtful whether my approach is indeed the most efficient/elegant one. Seems a bit clumsy to me...
x <- ddply(x, .(id), transform, year.min=min(year[!is.na(event)])) #identifies first event year
x1 <- ddply(x[x$year>=x$year.min,], .(id), transform, index=seq_along(id)) #creates counter years following first event; prior years are removed
x1 <- x1[order(x1$id, x1$year),] #sort
x1 <- ddply(x1, .(id), transform, growth=100*(value/value[1])) #calculate difference, however, based on first event year; this is wrong.
library(Interact) #i then merge the df with the years prior to first event which have been removed in the begining
x$id.year <- interaction(x$id,x$year)
x1$id.year <- interaction(x1$id,x1$year)
x$index <- x$growth <- NA
y <- rbind(x[x$year<x$year.min,],x1)
y <- y[order(y$id,y$year),]
Many thanks for any advice.

Answer 1

# Create a tag to indicate the start of each new event by id or
# when id changes
dat$tag <- with(dat, ave(as.character(event), as.character(id),
FUN=function(i) cumsum(!is.na(i))))
# Calculate the growth by id and tag
# this will also produce results for each id before an event has happened
dat$growth <- with(dat, ave(value, tag, id, FUN=function(i) i/i[1] ))
# remove growth prior to an event (this will be when tag equals zero as no
# event have occurred)
dat$growth[dat$tag==0] <- NA

Answer 2

Here is a solution with dplyr.
ana <- group_by(mydf, id) %>%
do(na.locf(., na.rm = FALSE)) %>%
mutate(value = as.numeric(value)) %>%
group_by(id, event) %>%
mutate(growth = value/value[1]) %>%
mutate(index = row_number(event))
ana$growth[is.na(ana$event)] <- 0
id value year event growth index
1 a 100.0 1950 NA 0.0000000 1
2 a 101.0 1951 NA 0.0000000 2
3 a 102.0 1952 NA 0.0000000 3
4 a 103.0 1953 NA 0.0000000 4
5 a 104.0 1954 NA 0.0000000 5
6 a 105.0 1955 X 1.0000000 1
7 a 106.0 1956 X 1.0095238 2
8 a 107.0 1957 X 1.0190476 3
9 a 108.0 1958 X 1.0285714 4
10 a 107.0 1959 Y 1.0000000 1
11 a 106.0 1960 Y 0.9906542 2
12 a 105.0 1961 Y 0.9813084 3
13 a 104.8 1962 Y 0.9794393 4
14 a 104.2 1963 Y 0.9738318 5
15 b 70.0 1970 NA 0.0000000 1
16 b 75.0 1971 NA 0.0000000 2
17 b 80.0 1972 NA 0.0000000 3
18 b 85.0 1973 NA 0.0000000 4
19 b 90.0 1974 NA 0.0000000 5
20 b 60.0 1975 Z 1.0000000 1
21 b 59.0 1976 Z 0.9833333 2
22 b 58.0 1977 Z 0.9666667 3
23 b 57.0 1978 Z 0.9500000 4
24 b 56.0 1979 Z 0.9333333 5
25 b 55.0 1980 W 1.0000000 1
26 b 54.0 1981 W 0.9818182 2
27 b 53.0 1982 W 0.9636364 3
28 b 52.0 1983 W 0.9454545 4

Answer 3

Try:
ddf$index=0
ddf$growth=0
baseline =0
r=1; start=FALSE
for(r in 1:nrow(ddf)){
if(is.na(ddf$event[r])){
if(start) {
ddf$index[r] = ddf$index[r-1]+1
ddf$growth[r] = ddf$value[r]/baseline
}
else {ddf$index[r] = 0;
}
}
else{
start=T
ddf$index[r] = 1
ddf$growth[r]=1
baseline = ddf$value[r]
}
}
ddf
id value year event index growth
1 a 100.0 1950 <NA> 0 0.0000000
2 a 101.0 1951 <NA> 0 0.0000000
3 a 102.0 1952 <NA> 0 0.0000000
4 a 103.0 1953 <NA> 0 0.0000000
5 a 104.0 1954 <NA> 0 0.0000000
6 a 105.0 1955 X 1 1.0000000
7 a 106.0 1956 <NA> 2 1.0095238
8 a 107.0 1957 <NA> 3 1.0190476
9 a 108.0 1958 <NA> 4 1.0285714
10 a 107.0 1959 Y 1 1.0000000
11 a 106.0 1960 <NA> 2 0.9906542
12 a 105.0 1961 <NA> 3 0.9813084
13 a 104.8 1962 <NA> 4 0.9794393
14 a 104.2 1963 <NA> 5 0.9738318
15 b 70.0 1970 <NA> 6 0.6542056
16 b 75.0 1971 <NA> 7 0.7009346
17 b 80.0 1972 <NA> 8 0.7476636
18 b 85.0 1973 <NA> 9 0.7943925
19 b 90.0 1974 <NA> 10 0.8411215
20 b 60.0 1975 Z 1 1.0000000
21 b 59.0 1976 <NA> 2 0.9833333
22 b 58.0 1977 <NA> 3 0.9666667
23 b 57.0 1978 <NA> 4 0.9500000
24 b 56.0 1979 <NA> 5 0.9333333
25 b 55.0 1980 W 1 1.0000000
26 b 54.0 1981 <NA> 2 0.9818182
27 b 53.0 1982 <NA> 3 0.9636364
28 b 52.0 1983 <NA> 4 0.9454545
29 b 51.0 1984 <NA> 5 0.9272727