cls_fast.m is comparunrol least squares which unrolls correctly the others don't work correctly, e.g. scraggle does least squares on the other axis, which gives wrong result. total least squares (TLS) may work somewhere in between also, ignore rcond warning; matrix ill conditioned but still works ok. may fix this by doing >256 equations in 256 unknowns. Jsum03.cls_fast.mat has the response function for skip=3 which is k=2 e.g. this is the answer when running with comparagram Jsum03. these files all generated in octave (e.g. .mat files not readable in matlab) to read in matlab copy to a text file and manually edit out header and tailer