apasto / decorrelate-grids

Given two 2D arrays (grids), perform a windowed linear regression between the two. Aim: removal of correlated component between two variables.
Apache License 2.0
8 stars 2 forks source link

Returned coefficients are some order of magnitude off #1

Closed apasto closed 2 years ago

apasto commented 2 years ago

When running de-correlations (in the dInSAR application, between topography and topo-correlated unwrapped phase), the returned c0 and c1 grids have values in the order of 10^5 and 10^2, respectively. Order of magnitude of A (unwrap) is 10^0, of B (topography) is 10^3. Performing the correction results in implausible values. The likely culprit: we are not computing A(B) regression, but B(A) instead.