2
$\begingroup$

I am trying to generate the parameters of a kumaraswamy distribution

https://en.wikipedia.org/wiki/Kumaraswamy_distribution

I have a mean and CV of a variable x

mean.x <- 3000
cv.x <- 0.1
sd.x <- cv.x * mean.x

I normalise the mean.x and sd.x by a max and min value as follows:

x.max <- 6000
x.min <- 0

norm.mean.x <- (mean.x - x.min)/(x.max - x.min)
norm.sd.x <- (1/(x.max - x.min)) * sd.x

What I want to do is to estimate (via some fitting procedure) the a and b parameters of the kuma distribution? Could anyone explain how this can be done in R? I guess maybe I can start with some default value of a and b, sample n values, find the mean and sd, compared it with the mean and sd I have to derive sum of squared errors and keep changing a and b parameter to arrive at that set that gives the lowest sse. For e.g.

  library(extraDistr)

  kuma.n <- rkumar(100, a = 2, b = 1)
  mean.kuma <- mean(kuma.n) 
  sd.kuma <- sd(kuma.n)

  ssq <- (norm.mean.x - mean.kuma)^2 + (norm.sd.x - sd.kuma)^2

and keep changing a and b. However, this seems like an endless iteration so was wondering if anyone has a better solution to this.

EDIT:

the above is a made up data.

$\endgroup$
1
  • $\begingroup$ The endless iteration might be a good job for the optim function in stats: see help(optim) or one of the many blogs on that in the net like r-bloggers.com/how-to-use-optim-in-r $\endgroup$
    – Bernhard
    Oct 24, 2019 at 12:13

1 Answer 1

1
$\begingroup$

This particular distribution has an ugly function describing the mean and variance. So I would propose using maximum likelihood estimation from sampled data. The likelihood function expresses the plausibility of different parameters given a data set. So I would do the following:

1) simulate a lot of data from your mean and cv

2) Fit the data to the Kumaraswamy Distribution using the following R package: https://www.rdocumentation.org/packages/VGAM/versions/1.1-1/topics/kumar the package is described here: https://www.stat.auckland.ac.nz/~yee/VGAM/

$\endgroup$
2
  • $\begingroup$ Thank you. The data in my example is just made up data. $\endgroup$
    – 89_Simple
    Oct 24, 2019 at 17:24
  • $\begingroup$ Can you please tell me how do I simulate data for a given mean and CV in your first point? $\endgroup$
    – 89_Simple
    Oct 24, 2019 at 17:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.