hello everyone
After I calibrated my average delay at roundabout using simulation random seed number of 42, then I tried to run the model 5 runs using different seed numbers, but I get different results. My question is, is this means that my model needs calibration or I have to run the model more times? #run #randomseed #calibration
Hello Shaza,
Different data yield different outcomes; there is no surprise there. However, if what you're trying to do is establish a summary statistic for your process's performance, you might consider one of the following:
1) Median performance of X trials.
2) Observed 5 and 95 percentile points of, say, 1000 or more trials.
3) To be conservative, you could report the worst observed performance across the set of trials. Thus, performance in future data sets would be normally expected to be at or above that level.
The developers of the software implementations of resampling statistics (Peter Bruce and Julian Simon), pretty much suggest that option #2 above is a good way to capture the "realistic" spread in obtained performance from a variety of trials.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running