I am currently using Mathematica for statistical analysis and would like to request the development of the following prediction functions for future versions. While some of these methods are already implemented, but it would be beneficial to refine and polish the existing implementation to improve their usefulness and funcionality. I appreciate your consideration of these suggestions and look forward to the potential future development of these functions in Mathematica.
- Partial Least Squares (PLS) - Projection to Latent Structures
- PLS is very popular in chemometrics and is expanding to PLS-Path
(similar to the SEM) - https://www.tqmp.org/RegularArticles/vol10-2/p200/p200.pdf
- PLS is very popular in chemometrics and is expanding to PLS-Path
- Parallel Factor Analysis (PARAFAC)
- Structural Equation Modeling (SEM)
- SEM is the most widely used method in education, social science, etc.
- https://content.wolfram.com/sites/19/2020/12/Oldenburg.pdf
- Correspondence Analysis (CA)
- CA is useful to analyze the proportions
- https://content.wolfram.com/sites/19/2010/09/Yelland.pdf
- Decision Tree (CART Regression and CART Tree)
- The Predict function includes decision tree features, but it
currently lacks a way to graphically display the decision tree
results. Adding visual representation of decision trees would make
this feature more user-friendly and effective for practical
applications.
- The Predict function includes decision tree features, but it
- Factor Analysis with Varimax Rotation
- The SingularValueDecomposition command is currently used for
Principal Component Analysis (PCA), but rotation methods,
particularly varimax rotation, are not implemented.
- The SingularValueDecomposition command is currently used for
- Independent Component Analysis (ICA) and Non-Negative Matrix
Factorization (NNMF) are already available in the Wolfram function
repository and more used cases would be valuable