Add mgx mtx function

[sterrettJD]

No description provided.

[sterrettJD]

specify in docstrings that feature.table is mtx

[sterrettJD]

addressed 478318a

[klterwelp]

fixed.vars=NULL

[sterrettJD]

addressed in 478318a

[sterrettJD]

Hey Kat, I think the main thing still needed is the last integration tests for mtxDE. I can add these if you'd like, but let me know if you'd like to do that instead.

What I mean by integration tests is just testing where a full example of mtxDE is run with a dna table too. I think all of the new intermediates have been tested, which is great!

[klterwelp]

An integration test has been on my to-do list! Since I wrote the code, is it better for me to do it? Or, is it better to have you test my code as an "outsider" point of view?

Which do you prefer? It'll be next on my to do list after simulations :)

[sterrettJD]

Hey @klterwelp, I went ahead and added a few basic integration tests. The lm tests pass, but the gamlss tests unfortunately fail due to some weird behavior with the GAMLSS models, where either the phenotype or MGX feature won't exist in the report.

That's because the parameter estimates are NA when broom.mixed::tidy goes to summarize them. I'm not quite sure why they're NA, but maybe bigger sample sizes would fix this? Or it's something about the distributions of the data in these tests?

Have you run into this? I can do some more digging soon.

[klterwelp]

That's weird! The gamlss method does give me estimates for "_mgx" and phenotype using the mtx2021 simulations. Maybe try running it with one of those to see if the behavior persists?

[sterrettJD]

Hey @klterwelp I found that it's mostly sample-size dependent. Ran some other examples where very small samples cause these to be NA, whereas larger samples don't have that issue. I just made more realistic dataset. It may be worth adding a check for this in the output, but I think it's okay for now