Using the example data I created below to emulate your data, we can convert our y variable into an indicator that is 1 when y isn't 0 and 0 otherwise. We can then use rleid() from the data.table package to create an identifier for each unique run of values in our y variable. If we remove the rows where y == 0, then we now basically have a group identifier for which rows correspond to which curve. We can then group_by() this identifier and get the within-curve maximum with summarise().
Note that this follows the OP in assuming that there are no value of exactly zero within a curve. In other words c(1,2,3,0,5,6,7,8,0,0,0) is two curves, not one.
library(dplyr)
library(ggplot2)
library(data.table)
set.seed(123)
# Example data
data <- data.frame(
x = 1:50,
y = c(runif(5), rep(0,5),
runif(5), rep(0,5),
runif(5), rep(0,5),
runif(5), rep(0,5),
runif(5), rep(0,5)))
# Show data
ggplot(data, aes(x=x, y=y)) + geom_line() + theme_bw()
# Get the maximum values between every run of zeroes
maxima <- data %>%
mutate(curve = data.table::rleid(y != 0)) %>%
filter(y != 0) %>%
group_by(curve) %>%
summarise(peak = max(y))
print(maxima)
#> # A tibble: 5 x 2
#> curve peak
#> <int> <dbl>
#> 1 1 0.940
#> 2 3 0.892
#> 3 5 0.957
#> 4 7 0.955
#> 5 9 0.994
# Get the mean
mean(maxima$peak)
#> [1] 0.9476986
