Inventories of the MARS Project of the IRSA that is briefly described. Some aspects concerning stability are discussed. Regression results seem to be more reliable if training pixels are chosen at random in a random subset of segments (elements of the area sampling frame). Some risk of bias is observed if segments with training pixels are included to compute the regression parameters.
No takes yet. Share an insight, caveat, or question.
Gallego et al. (1993) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: