Clustering method that is robust against inhomogeneous cluster sizes

Question

I have data clustered in various sets whose amount of member variables and occupied area varies, the former among multiple magitudes.

Here is an example:

Set A is very concentrated and has 100000 Elements. Set B is slightly more widespread and contains 1000 Elements. Set C is covers a comparably big area and has 10 Elements.

I am working on a one dimensional axis. Since I do not know the number of sets I want to obtain, I am looking for a clustering algorithm that returns a variable number of clusters.

So far, hirarchical clustering has given good results when working with two different magnitudes of cluster sizes, but has failed me with three magnitudes, as in the example above.

Answer 1

"Clustering" is mostly used for multivariate analysis.

In one dimensional data, you may achieve very good results by looking for minima in a variable width kernel density estimation.