Uncertainties of numerical results

Click For Summary
SUMMARY

The discussion centers on determining uncertainties in results derived from Monte Carlo simulations, specifically in the context of particle diffusion through a small hole in a confined box. Participants agree that performing multiple simulations and calculating the standard deviation of the results is a reliable method for estimating uncertainty. Additionally, it is suggested that if computational resources are limited, one can conduct simulations with smaller sample sizes and use curve fitting to extrapolate uncertainty as a function of the sample size, following the principle that uncertainty decreases with the square root of the sample size (1/sqrt(N)).

PREREQUISITES
  • Understanding of Monte Carlo simulations
  • Familiarity with Brownian motion concepts
  • Knowledge of statistical methods, particularly standard deviation
  • Experience with curve fitting techniques
NEXT STEPS
  • Research advanced Monte Carlo simulation techniques
  • Learn about statistical analysis in simulation results
  • Explore curve fitting methods for extrapolating data
  • Study the implications of sample size on uncertainty in simulations
USEFUL FOR

Researchers, physicists, and data analysts involved in computational simulations and statistical analysis, particularly those focusing on particle dynamics and uncertainty quantification.

Niles
Messages
1,834
Reaction score
0
How do people generally determine uncertainties on results that are based on Monte Carlo simulations? Take this fictive example, just so we have something specific to talk about:

I look at 106 particles, confined to a box. There is a small hole in one of the walls, and at some time t0 I am interested in knowing how many particles N have diffused out of the box. This can be simulated by a Monte Carlo approach (Brownian motion).

This number N will vary each time I perform the simulation, but it will converge the larger I make the initial sample. Nonetheless, I guess an uncertainty is still present - how can we determine that in general?

Thanks for input in advance.
 
Physics news on Phys.org
I'm not sure if there's a more elegant method, but I'd repeat the simulation multiple times and take the standard deviation of the results. Perhaps if it's too computationally expensive to do this with a very high N, you can repeat it for a few low N and extrapolate the uncertainty as a function of N using a curve fit? I assume it would be 1/sqrt(N)?
 

Similar threads

  • · Replies 6 ·
Replies
6
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 7 ·
Replies
7
Views
3K
Replies
0
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
Replies
11
Views
3K