tiny FIMS model

[iantaylor-NOAA]

The data1 input for FIMS included in /data/ has 19,080 rows which makes debugging harder because the output is similarly large (in spite of not being a particularly complex model).

I just created a tiny FIMS model with 3 years of data, 3 ages, and 3 length bins, which has only 60 rows for the input:
tiny_fims_model.r.txt.

Where would be the best place to put something like this to make it available for various uses (such as debugging #933)? Would it make sense to add it to /data/ as a .rda file even though the script to create it is pretty short?

[k-doering-NOAA]

Good idea!

Hmm, R packages has differing advice based on the format or purpose of the data, but sounds like if it is helpful for users, putting it in the data subfolder as an .rda makes sense. I think the script can go in the data-raw subfolder to preserve how the dataset was created.

[kellijohnson-NOAA]

I agree with Kathryn that this could be a data object in the package. We have data1 already but we could name it data_small or data_minimal? @iantaylor-NOAA is this something you are willing to open a PR for?

[kellijohnson-NOAA]

Are there any tests that use data1 and if yes can we change any of them to use the new smaller data frame? Is the smaller data frame big enough to produce reproducible results?

[iantaylor-NOAA]

@kellijohnson-NOAA, yes I can make a PR to add data_small (the name I like best of those you suggest).

We do have a bunch of tests that use data1 (see 10 files in tests here: https://github.com/search?q=repo%3ANOAA-FIMS%2FFIMS+path%3A%2F%5Etests%5C%2F%2F+data1&type=code), but I don't think it makes sense to change over to data_small immediately.

I'm not sure about lack of reproducibility but it probably makes sense to update the data_small values to slightly better match the expectations of a model fit to them rather than arbitrary stuff I just made up. I'll explore doing that as part of crafting the PR.