Re: Bimodal Gamma Distribution

The red line in yourfigure is NOT a Normal
distribution.
A Normal distribution, like a gamma distribution, has
only one
mode (peak). For on-line "literature" try
http://en.wikipedia.org/wiki/Normal_distribution
http://en.wikipedia.org/wiki/Gamma_distribution
http://www.causascientia.org/math_stat/Dists/Compendiu
m.pdf
and note that the expressions for the modes of the
Normal and
the Gamma are single valued. You may find the
"Continuous
Mixtures" section of Compendium.pdf interesting, but
working
with mixtures of distributions can be awkward.

Hope that helps,

A Lurker




The red curve looks like a mixture of two normals, while the green curve is that of a single gamma. A similar procedure used to fit a mixture of normals can also be used to fit a mixture of gammas or even a mixture of Weibulls.

Jack
.

Relevant Pages

  • Re: Shape and Scale parameters of Gamma distribution
    ... It is one way to _estimate_ the distribution parameters of the hypothetical underlying distribution from a sample of data. ... not to fit before generating random number. ... that's what GAMRND does. ... If you mean, "how well does the specific gamma distribution that I have estimated from my data model those data?", then that's the sort of thing that the second output of GAMFIT, or probability plots, or any number of other methods, helps to tell you. ...
    (comp.soft-sys.matlab)
  • Re: Using glmfit to estimate gamma mean AND variance
    ... I am trying to fit a regression model, ... distribution I'm trying to fit is gamma. ... Again, you are not estimating two parameters of a distribution, since this is a regression model. ...
    (comp.soft-sys.matlab)
  • Re: Using glmfit to estimate gamma mean AND variance
    ... I am trying to fit a regression model, ... distribution I'm trying to fit is gamma. ... its variance. ...
    (comp.soft-sys.matlab)
  • 3-parameter Gamma and Weibull distribution
    ... Do exist m-files that I could use to fit 3-parameter ... distribution: Gamma and Weibull ...
    (comp.soft-sys.matlab)
  • Re: Demonstration that least squares give maximum likelihood
    ... described by the pdf (the probability density function). ... I thought I understood that the normal distribution is derived by increasing the number of coins tossed in each experiment. ... analytical chemistry and the practice of discarding deviant observations ... tails than normals but nowhere as long as Cauchys. ...
    (sci.stat.edu)