re-structuring semtree.control()

[brandmaier]

This is a discussion on whether we want to improve the accessibility of the semtree.control-object based on ideas of @manuelarnold. He suggested to restructure as following:

• method: "LR" is default or "score"
• LR_tests: "maxLR" (default) or "naive","bonferroni","fair"
• score.tests: list(nominal = 'LMuo', ordinal = 'maxLMo', metric = 'maxLM') (default settings)
• fast_tree: FALSE (default), TRUE

I like the separation of the score vs LR and a second argument to specify how the LR tests are done. However, I suggest that instead of fast_tree as an argument, we provide a separate constructor for the control object, that is, you call semtree.control.fast() which then returns a control-object that has score.tests=list(nominal = 'LMuo', ordinal = 'WDM', metric = 'DM' and method="score".

[manuelarnold]

I like this idea a lot. It makes the original semtree.control function less crowded. By the way, we should either use a dot or an underscore in the argument names of a function.

[brandmaier]

I vote all underscore (LR_tests, score_tests, ...). For consistency, should it also be semtree_control ?

[manuelarnold]

I also prefer underscores because dots can be confused with S3 methods. I would go for semtree_control. Maybe we could keep a semtree.control function in the package with a warning that it is deprecated and will be removed in the future?