Bug 2513

Summary:	ParetoRandomVariable needs a "scale", not a "mean" attribute.
Product:	ns-3	Reporter:	Tommaso Pecorella <tommaso.pecorella>
Component:	core	Assignee:	Peter Barnes <pdbarnes>
Status:	RESOLVED FIXED
Severity:	enhancement	CC:	ns-bugs, tomh
Priority:	P5
Version:	ns-3-dev
Hardware:	All
OS:	All
Attachments:	New attribute and so on new patch - fixed a bug and the tests

Description Tommaso Pecorella 2016-09-30 19:22:43 UTC

Now that we have an attribute deprecation system...

The actual ParetoRandomVariable class has a "Mean" attribute, but it's limiting.

If we don't have a "Scale" attribute, the scale is calculated as:
double scale = mean * (shape - 1.0) / shape;

This works until shape is greater than one. If it's lower or equal to one... boom.

However, the problem is that the Pareto distribution parameters are the *Shape* and the *Scale*. It is totally legit for a Pareto distribution to not have a mean.

My proposal is to deprecate the Mean attribute and add a Scale attribute.
If the user sets the Mean, we have to check for the Shape to be greater than one (otherwise it's an error).
If the user doesn't set the Mean... well, all is fine.

Comment 1 Tommaso Pecorella 2016-10-01 06:26:08 UTC

Created attachment 2599 [details]
New attribute and so on

Comment 2 Tommaso Pecorella 2016-11-05 13:22:44 UTC

Created attachment 2653 [details]
new patch - fixed a bug and the tests

Comment 3 Tom Henderson 2016-11-05 17:02:50 UTC

(In reply to Tommaso Pecorella from comment #2)
> Created attachment 2653 [details]
> new patch - fixed a bug and the tests

OK by me; please update the 'changes to existing API' in CHANGES.html when pushed.

Comment 4 Tommaso Pecorella 2016-11-07 20:08:44 UTC

changeset 12397:6572761077e5