can any one suggest me how we can fit a distribution which is a sum of various standard distributions to given data or to a normalized histogram

I have a one-dimensional normalized arrays whose individual sum will be equal to 1. I would like to fit distributions(pdf) to each array. Can I know whether we can fit a distribution which is a result of various distributions like sum of two Gaussians, or the sum of a guassian and an exponential etc.. to the arrays. If yes, please let me know who can I do that. I am familiar with basic Matlab, but new to distributions fitting in matlab

Answer 1

[dpb]

See

doc cftool

for the interactive fitting tool; it includes builtin Gaussian with up to eight (8) terms; it doesn't have mixture models, unfortunately, but you can write custom functions for such.

There's a distribution-fitting specific tool in the Statistics Toolbox but I've never tried messing with it...

doc distributionFitter

Or, interactively, you can call cfit directly. The

Answer 2

[Jeff Miller]

You might be able to use Cupid . It provides a lot of ways to form new distributions by combining and transforming standard ones. No nice GUI, though.