[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ns] Re: ParetoII



  I've done a bit more investigation and have some more questions --

  The ParetoII distribution is not the Lomax distribution (as defined in
"Continuous Univariate Distributions, vol. I" by Norman Johnson and Samuel
Kotz).

  The sampling method for the Lomax distribution is

	x = (scale / uniform())^(1/shape) - C

  The sampling method for ParetoII (as used in ns-2) is

	x = scale * (1 / uniform()^(1/shape) - 1)

	  = (scale / uniform()^(1/shape)) - scale

  Interestingly, the sampling method for Pareto is

	x = scale / uniform()^(1/shape)

  So, ParetoII just subtracts scale from each sample generated by Pareto.
Why?

  Also, the scale value is different in Pareto and ParetoII.  Using the
distribution functions in ns-2, we can specify average and shape.  Then scale
is computed and is used to generate samples.

  For Pareto, scale = average (shape - 1) / shape

  For ParetoII, scale = average (shape - 1)

  Is there a reference for what ParetoII actually is and why it's used in the
webtraf simulations rather than Pareto?

Thanks,
Michele
--
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Michele Clark Weigle                        clark@cs.unc.edu
Department of Computer Science		    http://www.cs.unc.edu/~clark
University of North Carolina		    Sitterson #327
Chapel Hill, NC 27599-3175                  (919) 962-1909
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%