Identity basis

[BalzaniEdoardo]

In this PR I add an identity basis function and a basis function that capture raw history.

I added a few how-to guide that showcase the use of these basis for:

1. Including custom predictors in a composite basis.
2. Create a design from categorical variables.
3. Fit a coupled GLM.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 98.51852% with 2 lines in your changes missing coverage. Please review.

Project coverage is 97.27%. Comparing base (a510ef3) to head (89f279b).
Report is 60 commits behind head on development.

Files with missing lines	Patch %	Lines
src/nemos/basis/_identity.py	98.03%	1 Missing ⚠️
src/nemos/basis/_transformer_basis.py	97.61%	1 Missing ⚠️

Additional details and impacted files

@@               Coverage Diff               @@
##           development     #282      +/-   ##
===============================================
+ Coverage        96.13%   97.27%   +1.13%     
===============================================
  Files               34       35       +1     
  Lines             2642     2750     +108     
===============================================
+ Hits              2540     2675     +135     
+ Misses             102       75      -27     

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

[billbrod]

It looks like this PR is showing the changes from #280 in the diff, do you know why?

Otherwise, these looks generally good, though I think this means we're going to have to go through the existing how-tos ,etc to see where to use the new bases (at least in the head direction tutorial). Can we make an issue for that?

The new how-tos all looks like they match what they are in diataxis, i.e., very short and answering a single question. I think we'll need to go back through the existing ones and update / move them then, and I think we should consider changing the layout of the landing page for them: make the cards smaller, only show the top-level title, which should be very descriptive. This should also be a separate issue.

[billbrod]

why not do this one using the data from https://nemos--282.org.readthedocs.build/en/282/tutorials/plot_02_head_direction.html? then we can visualize the outputs (and point them to the full tutorial for more details)

[billbrod]

regardless, I think we should visualize the model coefficients we get

[BalzaniEdoardo]

i think it requires too much pre-processing for this session. I'll print the coefficients

[BalzaniEdoardo]

also, i would need more explaining about the data, maybe filter a subset of unit and samples... I think it is better to go straight to the point. If we had a sample data which has already a sub-population and a short duration we could load that (like the iris dataset for sklearn)

[billbrod]

Okay, I think this is fine, but it would be nice to think about whether there is a meaningful dataset we could use. because it's nice to plot the coefficients and have them look meaningful.

[BalzaniEdoardo]

you are right

[billbrod]

Suggested change

	You can directly pass a [`pandas.DataFrame`](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.html) to the
	`fit` method, as long as it can be internally converted to a floating-point array. The conversion is handled automatically,
	so no additional steps are required on your part.
	You can directly pass a [`pandas.DataFrame`](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.html) to the
	`fit` method, as long as it can be internally converted to a floating-point array (that means, for example, that you can't have a column whose values are strings). The conversion is handled automatically, by calling `design_df.values`,
	so no additional steps are required on your part.

[billbrod]

but it will fail if they try to do use a dataframe with non-numeric values anyway, right?

[BalzaniEdoardo]

?

It looks like this PR is showing the changes from #280 in the diff, do you know why?
No idea, I might have branched from set_shape_basis_method and this confused the history. I am not sure it makes sense

Otherwise, these looks generally good, though I think this means we're going to have to go through the existing how-tos ,etc to see where to use them. Can we make an issue for that?
done

The new how-tos all looks like they match what they are in diataxis, i.e., very short and answering a single question. I think we'll need to go back through the existing ones and update / move them then, and I think we should consider changing the layout of the landing page for them: make the cards smaller, only show the top-level title, which should be very descriptive. This should also be a separate issue.

I agree with both points