|
|
 | | From: | Yan ZHANG | | Subject: | a strange phenomena in discrete event simulation | | Date: | Mon, 03 Jan 2005 09:21:17 -0600 |
|
|
 | I first got analysis result:0.55. Then, run the simulation program and get
the simulation result: 0.53. So, there are big difference between the
simulation and the analytical results.
Again, I run the simulation program for 4 times and surprising all these 4
runs get the same result: 0.55. So, there is no difference.
I have tried this in three different computers and get the same phenomena.
Can you please comment on this, why there is big difference in the first run
while no difference in the later runs? any suggestions to resolve this?
--
Thanks and Best Regards
Yan ZHANG
http://www.nict.com.sg/zhang/
|
|
| | From: | A.L. | | Subject: | Re: a strange phenomena in discrete event simulation | | Date: | Mon, 03 Jan 2005 10:03:19 -0600 |
|
|
| On Mon, 03 Jan 2005 09:21:17 -0600, "Yan ZHANG"
wrote:
>I first got analysis result:0.55. Then, run the simulation program and get
>the simulation result: 0.53. So, there are big difference between the
>simulation and the analytical results.
>From simulation program you get the estimate of output value. This
estimate is random variable with such parameters as distribution and
dispersion. You cannot compare both numbers as "numbers", you can do
the comparison in statistical sense. See books by Kleijnen about
interpreting results of discrete event simulations.
A.L.
|
|
| | From: | A.L. | | Subject: | Re: a strange phenomena in discrete event simulation | | Date: | Mon, 03 Jan 2005 10:23:23 -0600 |
|
|
| On Mon, 03 Jan 2005 09:21:17 -0600, "Yan ZHANG"
wrote:
>I first got analysis result:0.55. Then, run the simulation program and get
>the simulation result: 0.53. So, there are big difference between the
>simulation and the analytical results.
>
>Again, I run the simulation program for 4 times and surprising all these 4
>runs get the same result: 0.55. So, there is no difference.
One more thing: when you do multiple runs, you must select different
"seeds" for random number generators. Otherwise you will always get
the same results. Simulation program is totally deterministic...
A.L.
|
|
| | From: | Nicolaas Vroom | | Subject: | Re: a strange phenomena in discrete event simulation | | Date: | Wed, 05 Jan 2005 13:20:13 -0600 |
|
|
| "Yan ZHANG" schreef in bericht
news:33sd4nF40dideU1@individual.net...
> I first got analysis result:0.55. Then, run the simulation program and get
> the simulation result: 0.53. So, there are big difference between the
> simulation and the analytical results.
Dear Mr ZHANG,
I have a great problem in understanding why you call that there
is a big difference between the two numbers 0.55 and 0.53
In my honnest opinion those two numbers are almost identical.
Ofcourse this is a matter of opinion.
The fact that a simulation program does not give the same
result as expected (analysed) is a common fact of life or science.
> Again, I run the simulation program for 4 times and surprising all these 4
> runs get the same result: 0.55. So, there is no difference.
A simulation program (for example) each time gives the same results
if the random numbers used are the same.
This can happen if the initial value of your random number each time
is the same.
If you use visual basic a random way to initialize your random number
generator is by using the following statement :
RANDOMISE(CLOCK)
Nicolaas Vroom
> I have tried this in three different computers and get the same phenomena.
> Can you please comment on this, why there is big difference in the first
run
> while no difference in the later runs? any suggestions to resolve this?
>
>
> --
> Thanks and Best Regards
>
> Yan ZHANG
> http://www.nict.com.sg/zhang/
>
|
|
| | From: | Yan ZHANG | | Subject: | Re: a strange phenomena in discrete event simulation | | Date: | Thu, 06 Jan 2005 06:38:08 -0600 |
|
|
| Thank you very much for your comments.
> I have a great problem in understanding why you call that there
> is a big difference between the two numbers 0.55 and 0.53
> In my honnest opinion those two numbers are almost identical.
> Ofcourse this is a matter of opinion.
>
> The fact that a simulation program does not give the same
> result as expected (analysed) is a common fact of life or science.
>
Here, I only give an example. Under other input, the difference may become
bigger and this may cause the reviewer doubt the analytical model.
> A simulation program (for example) each time gives the same results
> if the random numbers used are the same.
> This can happen if the initial value of your random number each time
> is the same.
Exactly, after initialized different seed in the random number generator, I
got the correct result.
> If you use visual basic a random way to initialize your random number
> generator is by using the following statement :
> RANDOMISE(CLOCK)
>
> Nicolaas Vroom
Thank you for your suggestions.
--
Best Regards
Yan ZHANG
http://www.nict.com.sg/zhang/
|
|
|
