• Should I normalize the data before using mixOmics?

Yes the data should be normalized and possibly pre-processed beforehand.  MixOmics will accept any type of data (raw, normalized) . Be aware that PLS methods will center and scale the variables by default. For the PCA methods, users will need to specify center = TRUE and  scale. = TRUE. rCCA takes the data as such (i.e. no transformation).

  • What should I do with missing values?

The function spls can handle missing values as the algorithm performs local regressions. However, if you want to do perform predictions, this won’t be possible if you have missing data. In that case you will have to impute the missing values in the testing data set using nipals. To replace missing values by 0 is not recommended. More radically, you can delete the entire rows with missing data.