I am looking for a generalized normal distribution with mean $\mu$ and variance $\sigma^2$ but with an additional parameter for the kurtosis to add fat-tails. I can't find it - can anybody help?
$\begingroup$
$\endgroup$
-
2$\begingroup$ Depends how you want to generalize, but for instance the Student t distribution becomes the standard normal distribution for high degrees of freedom. $\endgroup$ – Raskolnikov Dec 21 '10 at 10:33
-
$\begingroup$ @Raskolnikov: Ideally I want to let the parameters for mean and variance stay the same but having an additional parameter for kurtosis. $\endgroup$ – vonjd Dec 21 '10 at 11:00
-
1$\begingroup$ That's can be obtained with any distribution with at minimum two free parameters. Just relabel the parameters so that the mean and variance are $\mu$ and $\sigma^2$. Your constraints are too weak to give anything interesting. $\endgroup$ – Raskolnikov Dec 21 '10 at 11:22
-
$\begingroup$ I cannot follow: The other constraint is that when the third parameter is e.g. 3 (or 0) we are back to the normal distr. $\endgroup$ – vonjd Dec 21 '10 at 11:45
$\begingroup$
$\endgroup$
I think Pearson type VII distribution might be what your are looking for. You will need to reparametrize it, but wikipedia page has the necessary formulas.
-
$\begingroup$ this was really very helpful - Thank you. Do you know if there is also a log-pearson type vii distribution? I can't find any references in google... $\endgroup$ – vonjd Dec 21 '10 at 16:36
-
1$\begingroup$ @vonjd, you're welcome. Thank helpful people who mentioned this distribution in kurtosis wikipedia page, that is how I found it. If by log-pearson you mean distribution of random variable whose log is pearson VII, you can derive its distribution function similarly to log-normal. Hence no readily available references. $\endgroup$ – mpiktas Dec 21 '10 at 20:47
