I got two multidimensional datasets, X and Y. I thought I build the model, which explains the relationship between two datasets, using canonical correlation analysis (CCA). The first correlation between two components(canonical variates) is 0.87. I think which is not so bad.

The problem is I have to do regression. For example, how does Y change what if a new X' comes in. So I need to get new components of Y, and then reconstruct it like the original Y.

I am not good at this area, so I can't do anything right now.

For more information, originally the dataset X has only categorical variables, so I vectorized it into binary (= One-hot encoding). And Y has 6 variables.

It seems to be related to Reduced-Rank Regression or Partial Least Squares as far as I know. And if you give me some examples of this using R, it would be much helpful and much appreciated.

  • $\begingroup$ Welcome to the site. What is your question? If you ask about code I think it is off-topic for the site, if your question is about how valid is your approach, it could be on topic $\endgroup$ – llrs Mar 12 '19 at 10:51
  • $\begingroup$ The approach is the main question.I just want to know how valid my approach and share idea of other people. As you can read, R code is just extra if it is possible. $\endgroup$ – kiddwill Mar 28 '19 at 10:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.