Ip heckman 2nd trial

[johamunoz]

[hanneoberman]

This is a second attempt to include the IPD Heckman imputation function (see #485)

[hanneoberman]

Check whether conditional import can be used?

[stefvanbuuren]

Johanna and Hanna, thanks for your contributions.

The PR is technically correct, but I hesitate to merge this version into master. The problem is that it would add many new dependencies to mice. To check the package, I needed to install 24 additional packages into my already sizeable library. Some of these dependencies are pretty large, e.g. VGAM or copula. Merging these into master would substantially increase checking time and complicate maintenance.

I see two possible routes to progress:

1. Rewrite the code of mice.impute.2l.heckman() so that it uses libraries already available to mice;
2. Create a separate package with specialised mice.impute.xxx() functions.

Option 1. might require substantial double work. Feel free to reopen if your choose route 1. I suggest opting for possibility 2. There are several packages, e.g. miceadds and micemd, that offer additional functions on top of mice. You could follow a similar path and exert direct control over your contributions.

Trust you understand my concerns.

[johamunoz]

Thank you very much to you! We understand the situation, and the implications of installing and monitoring an imputation method with so many dependencies. We will consider the options you raised. Thank you again, Johanna