Not my domain, so apologies if some of my terminology / explanation is not quite right.

mrregister claims that l2 is the default robust estimator. However it seems to me that in the absence of that command-line option being explicitly specified, no "robust" estimator will be used (selection; invocation). The "default" robust estimator predicated on a robust estimator being used doesn't seem to come into play.

This is unlike population_template, which claims that the absence of a robust estimator is the default. This seems to be faithful given that linear_estimator will contain None if that option isn't specified, which will result in mrregister not being provided with the rigid_metric.diff.estimator option.

So is the issue purely in the mrregister help page? Ie. it should say that no robust estimator is the default? Or is the intended default behaviour different?

(I'm looking to make "none" an explicit option choice in both of these cases; but I want to make sure first that my understanding is correct)