CODEWITHSUNDEEP
r
stats_noob
stats_noob

Reputation: 5925

Integration Error: length(lower) == 1 is not TRUE

I am trying to numerically integrate a probability distribution over a range of values. This should look something like this. For example, I would like to do this with the Exponential Probability Distribution Function (https://en.wikipedia.org/wiki/Exponential_distribution, http://homepages.math.uic.edu/~jyang06/stat401/handouts/handout8.pdf):

## define the integrated function (for lambda = 2)

integrand <- function(x) {2 * 2.718^(-2*x)}

upper = seq(0, 5, by = 0.01)
lower = 0

data = data.frame(lower,upper)
data$integral = integrate(integrand, lower = data$lower, upper = data$upper)

Unfortunately, I got this error:

Error in integrate(integrand, lower = my_data$lower, upper = my_data$upper) : 
  length(lower) == 1 is not TRUE

This is strange because the integrate function works normally otherwise:

> integrate(integrand, lower = 0, upper = 0.01)
0.02020132 with absolute error < 2.2e-16

> integrate(integrand, lower = 0, upper = 0.06)
0.127496 with absolute error < 1.4e-15

Thank you!

Upvotes: 1

Views: 1166

Answers (2)

Jenna Grimshaw
Jenna Grimshaw

Reputation: 53

Just in case anyone ran into a similar problem that I had with this error... I was trying to run an integrate function on my data:

sei <- function(t = 1) exp(-t)/t # standard exponential integral
alpha <- vegan::fisherfit(data)$estimate # alpha estimate
N <- sum(data) # number of observations

rank.logseries <- sapply(data, function(x) {
    logeqn <- x * log(1 + alpha/N)
    f<-integrate(sei, logeqn, Inf)
    fv<-f[["value"]]
    alpha*fv
})

And I got the same error. I realized the issue was my data was a single column dataframe. Once I clarified this, the integrate worked perfectly.

sei <- function(t = 1) exp(-t)/t # standard exponential integral
alpha <- vegan::fisherfit(data)$estimate # alpha estimate
N <- sum(data) # number of observations

rank.logseries <- sapply(data[,1], function(x) {
    logeqn <- x * log(1 + alpha/N)
    f<-integrate(sei, logeqn, Inf)
    fv<-f[["value"]]
    alpha*fv
})

Upvotes: 1

Kra.P
Kra.P

Reputation: 15143

Try this

data$integral  <- apply(data, 1, function(x) {integrate(integrand, lower = x[1], upper = x[2])$value})

head(data)
  lower upper   integral
1     0  0.00 0.00000000
2     0  0.01 0.02020132
3     0  0.02 0.04081069
4     0  0.03 0.06183635
5     0  0.04 0.08328672
6     0  0.05 0.10517036

Upvotes: 1

Related Questions