New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
ENH: linalg helper: orthogonal complement of matrix #3039
Comments
bump I think something like this would be useful for transforming spline basis functions in #5296 If we have some spline basis, and we want to remove some subspace from it.
in two steps we could do this with
The orthogonal_complement above sounds like it's just getting the "null" space. We want something similar to CCA that includes the projection on a set of conditioning variables. #5296 adds a |
see also #4992 for organizing linalg tools |
another case, I don't know whether that is handled by the above reduced rank partial projection
the pca function can do this, but we need to know the keepdim in advance
|
@josef-pkt great code love it |
@josef-pkt are you a professor? |
if so sorry for my language, the code is great |
(I edited your comment a bit to keep the language here clean.) Thanks, it's good to here that I'm not the only one that likes this kind of code. The code itself is pretty simple, but it's difficult for someone who is not a linalg expert which I'm not. |
docstring missing, e.g. what is |
great code! just came back to see |
checking my local files ex_gam_bs_autos_cleaned.ipynb includes a section on score_test using partial_project and pca aside: I didn't find a GAM notebook in the doc examples |
(parking another helper function, so it doesn't get lost)
Written to understand orthogonal complement which is heavily used in VECM/cointegration.
Should also show up in canonical correlation, reduced rank regression and similar.
Also looks similar to reparameterization to include constraints (which I wrote without understanding the math).
general story: econometrics books often refer to things without telling how to compute it.
Eg.
constraints: w.l.o.g. we only consider this constraint, because the other forms can be obtained by reparameterization/transformation".
Related to here: normalization of cointegrating vectors to [eye, beta_2] requires a sequence of equations so that first variables with coefficients normalize to eye include elements of all cointegrating vectors. (But authors like Lutkepohl don't say how.)
My guess, not verified, is that below this will show up in the inv of the first part during normalization.
I need an example to check this.
The text was updated successfully, but these errors were encountered: