Return multiple p-values from tests.

[pitakakariki]

Save time by running multiple tests per iteration, instead of redoing all that simulating and refitting for each test.

Could be:

• multiple tests
• test multiple variables

This will help with precision analysis down the line.

[palday]

I've been working for this for a while in the simultaneous branch on my fork, but there do appear to be some subtelties in using all the p-values returned from a certain test simultaneously (powerSimSimultaneous) and running multiple tests per simulation run (powerSimMultiple). There are some obvious limits to which tests support the previous omnibus method, but there seem to be some other weird edge cases that pop up from time to time where the one method works but the other doesn't for reasons I haven't yet figured out.

[pitakakariki]

Need to do some research before implementing this. Might not be as simple as it looks, because if we're testing multiple covariates they'll all need their own MDE, but we might be more interested in the power when the other covariates are at their expected sizes. This might be more relevant in glmm's than in lmm's.

[pitakakariki]

More discussion here https://github.com/palday/simr/issues/5 and here https://github.com/palday/simr/issues/4