GNOME Bugzilla – Bug 144717
Any way to add positive skew to randnorm?
Last modified: 2009-10-11 17:15:16 UTC
Can't seem to figure out if this is possible. I still want something that goes negative so the other pdfs aren't adequate. I'm using gnumeric 1.2.8 on the latest FC2
If you add a positive skew to randnorm then it isn't normal anymore. The usual procedure to get a differently shaped random distribution is to take randnorm and transform the result before using it.
If you are more specific in which kind of pdf you would like to see added we can surely accomodate you. THis request to date is unfortunately much too vague.
Go this link and click on "show the density function" http://azzalini.stat.unipd.it/SN/plot-SN1.html Something like that. More info on it available here: http://azzalini.stat.unipd.it/SN/
Thanks, I guess you are looking for a random number generator for the univariate skew-normal distribution. I would think we can add that one.
Yes. I should have been clearer earlier. Guess I was thinking it would have been an option in randnorm but can see that you might want it separate altogether.
We should probably also include the density function, cumulative distribution function etc of the skew normal (and skew t etc. distributions.) THe code for these are available in R (licensed via GPL 2 or later): http://r-forge.r-project.org/plugins/scmsvn/viewcvs.php/pkg/fGarch/inst/unitTests/runit4A.R?rev=1871&root=rmetrics&view=markup I am wondering whetehr we can use the "generate" script in fn-r to include this code. Morten?
Sorry, I copied the wrong link. It should have been: http://r-forge.r-project.org/plugins/scmsvn/viewcvs.php/*checkout*/pkg/fGarch/R/dist-snorm.R?rev=4174&root=rmetrics
And that should have been LGPL
Translating that to C is probably less effort to do by hand than using one of my hacky scripts. Also, the numerical quality of that code is -- at a first look -- less than the quality of the C code.
RANDSNORM and RANDSTDIST have been added. They will be in 1.9.15.