Re: [Q] simulation via ns

[N Swaminathan <swami@veena.cs.iitm.ernet.in>]

Hi Kim,

	What ur advisor said is right. You should run the
simulation many times but very important ! with different
seed ! that's it. Determine the confidence interval of the
parameters of interest and keep the confidence interval
greater than 95 %. If these terms are greek and latin to 
you, best is to go through some elimentary books on
Statistics !

-swami

On Fri, 14 Nov 1997, Eunsook Kim wrote:

> Dear all,
> 
> I simulated my retransmission algorithm via ns based on SRM with
> variable topologies, number of nodes, value of delay net, and packet
> size. I have thought how many times I simulated is not important in ns,
> but my advisor told me I'd better acquire the results from more than 500
> times running my program. I think it is very silly question, but I'd
> like to know if it is necessary. He said that I can get trusted results
> only if I make my results by selecting my data among the large amount of
> simulation results with following a statistical distribution such as
> poisson distribution. Actually, I don't know such a mathematical concept
> well. I just thought ns provide such thing already.
> Please let me know how much ns provides such function. In addition, I'd
> like to know the meaning of ns run time. I mean how I can decide trusted
> run time for a algorithm.
> Looking forward to hearing from you soon.
> Thanks in advances.
> eunsook
> (eunsook@cs.sookmyung.ac.kr)
> 
>