Availability of “max-min detector” in gemmr?

[pauldhami]

Greetings,

Firstly, thank you for this tool.

I am working with a dataset in which my current sample to raw feature ratio is approximately 3:1 (neuroimaging) and 10:1 (behavioural). I wish to apply PCA before performing CCA, and in your manuscript, you describe evaluating the number of PCA components other than 100 as used in the main manuscript. In the Supplementary Materials Figure 10, you mention:

"Here we chose the number of principal component with a “max-min detector”. As the algorithm provided multiple values for the optimal number of components pX (neuroimaging data) and pY (behavioral and demographic data), we selected here the pair that minimized pX + pY."

I was able to follow your code with my own data using (https://gemmr.readthedocs.io/en/latest/analyses_from_paper/empirical_data_how_many_pcs.html), but now I am unsure how to choose the optimal number of components for my X and Y matrices. I was wondering if the above mentioned "max-min detector" algorithm is included in gemmr as well?

Thank you, Paul

[pauldhami]

Apologizes, I just found the function: gemmr.model_selection.max_min_detector