Add data_mask to InjectiveData

[phajy]

This pull request adds a new field data_mask to the InjectiveData struct. The data_mask is a BitVector that allows for masking specific data points in the codomain array. This feature is useful for filtering out certain data points during analysis or modeling. The data_mask is initialized with all true values by default.

[phajy]

I think adding a data_mask to InjectiveData will be helpful. However, I'm not certain of the easiest way to update make_model_domain to create a suitable "high-low" energy grid. Do we need an equivalent of the objective_transformer in, e.g., spectraldata.jl to take care of things?

This is the bit of code that needs work:

https://github.com/phajy/SpectralFitting.jl/blob/ef6bd6627cb06578a10a45072ebd07f80a4aeb13/src/datasets/injectivedata.jl#L25-L31

Maybe we need to update objective_transformer?

https://github.com/phajy/SpectralFitting.jl/blob/ef6bd6627cb06578a10a45072ebd07f80a4aeb13/src/datasets/injectivedata.jl#L49

[fjebaker]

Hi Andy! This looks like a really good start!

Maybe we need to update objective_transformer?

Yes -- as you have no doubt noticed, if we mask the obective (co)domain, then there will be a discontinuity between the domain we evaluate the model on and the domain we want to calculate the fit statistic on. The role of the objective_transformer is to transform the objective (e.g. flux) of the model onto the same domain as the data that we have (optionally to do things like response folding etc.).

So in this case, for injective model and data, we can just have a transformer that applies the mask to the model as well, and our work is done.

For the cases we were discussing in #70, we will need to handle the transformation somehow, but we can do that in a seperate pull request. Let's focus on getting your changes in first, and then iterate from there.

[fjebaker]

objective_transformer needs modifying, and ideally some test cases added

[phajy]

Oops - thanks for adding the qualifiers.

[phajy]

When making the objective variance we have:

SpectralFitting.jl/src/datasets/injectivedata.jl

Line 39 in 1941d43

if !isnothing(dataset.domain_variance)

I think this should be codomain_varaiance so I've updated it in the PR.

[fjebaker]

I think this should be codomain_varaiance so I've updated it in the PR.

Well noticed! Thanks!

[phajy]

Third time lucky with the test case? Shouldn't cut and paste into a Julia environment where all the undeclared variables are already declared! Assuming it passes this time, this PR can probably be closed.

[codecov-commenter]

Codecov Report

Attention: 11 lines in your changes are missing coverage. Please review.

Comparison is base (2ff89dd) 67.11% compared to head (56f0ba4) 66.90%.
Report is 5 commits behind head on main.

Files	Patch %	Lines
src/model-data-io.jl	66.66%	11 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #71      +/-   ##
==========================================
- Coverage   67.11%   66.90%   -0.21%     
==========================================
  Files          40       40              
  Lines        1952     1988      +36     
==========================================
+ Hits         1310     1330      +20     
- Misses        642      658      +16     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.