I am looking at the documentation for the current development version (ee423b). I see the jacobian flag for the optimize function is described in two places, R/model.R and man/model-method-optimize.Rd.
In both cases, the jacobian flag is described in terms of "MLE" and "MAP" which differ from the broadly-understood meanings of the terms; in particular, those terms don't correspond to MLE and MAP as applied to the user-defined model.
My advice is to just expunge the stuff about "MLE" and "MAP" in the text describing the jacobian flag. The remaining text about optimization in the constrained vs unconstrained space covers what needs to be said.
(1) There is a reference to the CmdStan documentation, but the current version (https://mc-stan.org/docs/cmdstan-guide/optimize_config.html#jacobian-adjustments) describes jacobian without using those terms. So it isn't necessary to bring in those terms for consistency with CmdStan.
(2) It is pretty deeply confusing to any user to talk about MLE and MAP in nonstandard terms. There have been a few discussions in which users try to understand what's going on, e.g. https://discourse.mc-stan.org/t/what-does-jacobian-true-do-in-model-optimize-data-data-list-jacobian-true/36867/12 .
FWIW, and thanks for your attention to this.
I am looking at the documentation for the current development version (ee423b). I see the
jacobianflag for theoptimizefunction is described in two places,R/model.Randman/model-method-optimize.Rd.In both cases, the
jacobianflag is described in terms of "MLE" and "MAP" which differ from the broadly-understood meanings of the terms; in particular, those terms don't correspond to MLE and MAP as applied to the user-defined model.My advice is to just expunge the stuff about "MLE" and "MAP" in the text describing the
jacobianflag. The remaining text about optimization in the constrained vs unconstrained space covers what needs to be said.(1) There is a reference to the CmdStan documentation, but the current version (https://mc-stan.org/docs/cmdstan-guide/optimize_config.html#jacobian-adjustments) describes
jacobianwithout using those terms. So it isn't necessary to bring in those terms for consistency with CmdStan.(2) It is pretty deeply confusing to any user to talk about MLE and MAP in nonstandard terms. There have been a few discussions in which users try to understand what's going on, e.g. https://discourse.mc-stan.org/t/what-does-jacobian-true-do-in-model-optimize-data-data-list-jacobian-true/36867/12 .
FWIW, and thanks for your attention to this.