Jump to content

Randomness during experiments


Recommended Posts

Hi All!


I would like to evaluate the system I build with multiple parameter settings using experiments. Although I enabled randomness, some results among different experimental setups are exact copies. To be specific: the source station generates the same ('random') time sequence of generated entities for all experimental setups (time sequences for replication 1 in Scenario 1 and Scenario 2 are the same). Although there is randomness among replications within one exeperimental setup (time sequences for replication 1 and 2 in Scenario x are different), I would like 'more' randomness.

Does one of you have a suggestion?


Thanks in advance!


Wouter

Link to comment
Share on other sites

Attaching your model would be helpful so we can see what you are doing.

Are you specifying random number streams? Have you incorporated any replication-specific logic in your model?


To greatly oversimplify... you could think of every replication 1 uses seed 1 for every scenario. Every replication 2 uses seed 2 forevery scenario. ... So if you set up 5 identical scenarios (e.g. the controls are the same and the reps required are the same) then they all use the same pseudo-random numbers and you will get identical results for each rep of each scenario. But each rep 2 will be different than rep 1 because it is using a different seed.


As soon as you change anything from scenario 1 to scenario 2, then you will probably cause the streams to be pulled in a different order and then generate different results.

Link to comment
Share on other sites

×
×
  • Create New...