1

This question might seem silly at first read. But in my research, first I have done mathematical analysis and simulation and provided the results in section I. Later, in section II, based on Sec. I, I did simulation in which nodes are distributed over 2D plane and some parameter is calculated over the mobile-nodes (parameter is "transmission rate" to be precise).

Now, since every run of simulation will distribute node on a different location (as in a real-world environment where mobile-users can be located anywhere), hence I will get a different result in each simulation run.

How can I present this result in my paper now?

|improve this question
  • 1
    Technically speaking any simulation based on MC has different every times. The point is to present the part that reproducible / find the conclusion that is relevant to majority the random scenarios – Greg Jun 2 '18 at 5:38
  • 1
    You need to show enough to prove that your results are not achieved by mere luck. Reproducibility is not as much as achieving the exact same results, but rather about being able to repeatedly support your conclusions. – Yet Another Geek Jun 2 '18 at 15:51
  • 1
    Don't MC has to converge to a certain result? – Oleg Lobachev Jun 2 '18 at 16:09
  • Thanks for the comments. btw, what does MC means? – Sjaffry Jun 3 '18 at 10:50
  • @Sjaffry Monte Carlo – fedja Jun 4 '18 at 3:09
4

  1. It is important to run multiple indepemdent replications and statistically analyze the results rather than simply running the simulation once and reporting that result.

  2. It is also important to save the random number seeds that were used so that the simulation results can be replicated.

|improve this answer
  • Thanks for your reply. Point (2) was what was in my mind. Thanks for confirming that. I was thinking to run instead of multiple runs randomly, I should fix the seed and then demonstrate my result with this data. About point (1) can you please elaborate a bit how I can statistically analyze this result? – Sjaffry Jun 2 '18 at 4:13
  • If it can help, x-axis contans the rate, y-axis contains the probability of achieving that rate – Sjaffry Jun 2 '18 at 4:15
  • thx for the reply. I got what I want with your response. Thanks – Sjaffry Jun 3 '18 at 10:50
1

Why not take the average of a number of runs. In my opinion, if you use the seed value and the result cannot be replicated except by it, please mention the seed value and preferably justify why. Another option is to make your source code available with a proper documentation of how your results were gotten.

|improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.