Related Experiment Videos
A Metropolis Monte Carlo implementation of bayesian time-domain parameter estimation: application to coupling
1Department of Chemistry, Yale University, New Haven, Connecticut, 06511, USA.
Summary
Bayesian statistics enables parameter probability estimation from noisy signals. This study implements Bretthorst
Area of Science:
- Statistics
- Computational Science
Background:
- The Bayesian perspective allows assigning probabilities to unknown parameter values.
- Estimating parameters and their distributions from noisy data is crucial in many scientific fields.
Purpose of the Study:
- To implement Bayesian parameter estimation using the Metropolis Monte Carlo sampling algorithm.
- To assess the accuracy and realism of error estimates derived from this method.
Main Methods:
- Utilized the Metropolis Monte Carlo sampling algorithm for parameter and error estimation.
- Implemented Bretthorst's Bayesian parameter estimation formalism.
- Applied the method to estimate coupling constants from antiphase doublets.
Main Results:
- Demonstrated realistic error estimates through Monte Carlo sampling.
- Successfully estimated coupling constants from both synthetic and experimental data.
- Showcased the flexibility of the method with minimal assumptions on posterior distribution shape.
Conclusions:
- The implemented Bayesian approach with Monte Carlo sampling provides realistic parameter and error estimates.
- This method accurately estimates coupling constants, even from complex spectral data.
- Prior knowledge can be readily incorporated into the Bayesian framework.