Update to -domin- on SSC: v3.5

[Joseph Luchman]

With many thanks to Kit Baum, -domin- has been updated to version 3.5 on SSC.

New features:

• Built-in linear regression/-regress- function for speedier dominance analysis/DA

Invoking -domin- with no -reg()- or -fitstat()- entries now, like before, conducts linear regression-based DA but now uses a covariance-based method similar to that used by the "lmg" method in the R package `relaimpo`.

This significantly speeds up the computation of all dominance statistics by avoiding repeated calls to -regress-.

Invoking -domin- with no -reg()- or -fitstat()- entries no longer produces warnings. Both or neither -reg()- or -fitstat()- must be filled - this is a change from pre-3.5's behavior.

• Updates to -epsilon- method

Support for weights and update to the way it accesses the data to reduce memory needs.

Some big user-visible changes:

• Minimum supported version is 15

The version associated with the Stata Journal article remains fixed at Stata 12.1

• New option -noesampleok- and requirement that underlying command has e(sample)

Invokes -domin- pre-3.5's sample size determination method. Otherwise, -e(sample)- from estimation command is used as it better to ensures correct sample size is used.

It is now an error to use -domin- without specifying -noesampleok- when the underlying command does not return e(sample).

• -mi estimate- support moved to wrapper

-mi- and -miopts()- arguments depreciated and will produce errors. MI support is moved to wrapper program -mi_dom- shipped with -domin-

• -domme- now sub-module of -domin-

No longer a separate package to unify development process. SSC package -domme- removed.

Other changes:

• -domin- no longer -moremata-

-domme- remains dependent; to be removed in a later version.

• most of command migrated to Mata

Allowed for the use of the built-in linear regression method and the possibility of other custom methods in the future.

• Argument check changes

Many errors' conditions changed/updated for updates to command

Please do file issue reports on -domin-'s GitHub (preferred) or email me directly if you encounter bugs or have requests for development.

- joe