How to generate synthetic data with specific spatiotemporal correlation

by JoshD   Last Updated September 18, 2018 11:19 AM

My dataset represents a field evolving over time, so has dimensions [X,Y,T].

I would like to generate synthetic data with the same autocorrelation structure and spatial correlations as the real data (to validate the significance of some Hidden Markov clustering) but I am not sure how to do this.

For each single gridpoint with dimension T, I can get the desired autocorrelation through Fourier phase randomisation. However a naive attempt to then apply the same method in X and then Y afterwards gives incorrect results.



Related Questions



Testing data for regular oscillations

Updated April 29, 2015 21:08 PM



Metric for spatial auto cross correlation?

Updated July 31, 2017 12:19 PM