Fit outcome model tests

[mvankessel-EMC]

I've added tests for each parameter for fitOutcomeModel().

There are a several things that I'd like to clarify, if the behavior is intended:

1. interactionCovariateIds, excludeCovariateIds, and includeCovariateIds currently allows negative values.
2. profileGrid and profileBounds are allowed to be NULL / NA simultaneously.
3. profileBounds may be <= 0.
4. prior and control get silently not used, when useCovariates is FALSE, without any message or warning.

[schuemie]

Great work! To answer your questions:

1. interactionCovariateIds, excludeCovariateIds, and includeCovariateIds currently allows negative values.

In theory one could create negative covariate IDs. Currently nothing would break if you did.

2. profileGrid and profileBounds are allowed to be NULL / NA simultaneously.

Yes. I sometimes use this to avoid generating a likelihood profile (computing a likelihood profile can be computationally expensive).

3. profileBounds may be <= 0.

Yep. Bounds are on a log scale (as documented), so can be negative.

4. prior and control get silently not used, when useCovariates is FALSE, without any message or warning.

Yes, we only allow shrinkage on covariates that are not the effect of interest. If there are no covariates that are not the effect of interest, you can specify prior and control but they have no effect. Not sure if we should message about that.

[mvankessel-EMC]

Ok, then feel free to merge.