1

Consider a factory simulation. Y=how long does it take to produce 500 widgets. If we run the simulation once we would have the simulated observation Y1. If we wanted to understand Y1 is an outlier, what we would like to do is repeat the simulation and CONTINUE using the same with the same RNG so that we have an array of observations [Y1, Y2, Y3 ... Yn]. Why?? So that we can apply statistics to the simulated observations.

When simulating a system, it is often necessary to replicate simulations. This means running the same simulation 20+ times consecutively with the same RNG (Random number Generator) stream and logging each observation (output of interest for each simulation run).

After each simulation run is complete, the clock and every metric is reset: all entities are removed from the system. The RNG is NOT reset between simualations. The RNG is not reset after each simulation completion because this would result same simulated observations in each simualtion.

The simulation output for each run are logged for statistical evaluation.

Questions

1) Does Simevents already have this built-in and I need to sharpen my Googling skills?

2) Is there an example of replication with a simple model using Simevents?

gatorback
  • 1,351
  • 4
  • 19
  • 44
  • Is it se_setseeds, se_getseeds and/or se_randomizeseeds you are looking for or am I misinterpreting your question? – fjut Sep 12 '16 at 17:39
  • are you looking for something built-in or would a Matlab script to replicate simulations suit you ? – titus Sep 27 '16 at 15:05
  • Open to all good ideas: references and examples are appreciated. Thank you – gatorback Sep 27 '16 at 16:48
  • @fjut Good question: the se_XXXX commands are not what is called for in this case. – gatorback Oct 11 '16 at 14:02
  • @titus: built-in is preferable, but if not then a script could do the job. In my experience other very old simulation packages have this capability built-in so it is very well possible that this is built-in – gatorback Oct 24 '16 at 14:03

0 Answers0