Don't know where your data is from? Bayesian modeling for unknown coordinates
5 hours ago
- #Bayesian Modeling
- #Spatial Statistics
- #Gaussian Processes
- Spatial probability models are particularly useful in mining for predicting mineral concentrations at unobserved locations using correlated sample data.
- The example uses uranium and vanadium concentration data from Walker Lake, showing how Gaussian processes can be adapted when data locations are noisy.
- A Bayesian model incorporates latent true coordinates with Gaussian noise, allowing inference via Monte Carlo methods despite computational challenges.
- Increasing coordinate noise leads to larger posterior uncertainties in location estimates, but key spatial features remain identifiable in predictions.
- Comparison with naive kernel smoothing shows the GP model retains spatial structure better under location errors.