Many statistical software ask whether to standardize data or no:

  1. What is a general rule to when data should be standardized?

  2. Do we standardize categorical variables?

  3. Is there a difference in how standardization effects or is interpreted in different regressions (linear, logistic, cox)

  • $\begingroup$ Data standardization is needed when there is distance computation, such as Euclidean, involved between the observations, or there is product of X and X_transpose computation. For instances, ridge regression needs the data to be normalized (or standardized) which involves product of X and X_transpose computation. Most common approaches for standardization are mean-std standardization and min-max standardization. Categorical variables can be replaced via on-hot encoding. stats.stackexchange.com/questions/186031/… $\endgroup$
    – prashanth
    Jun 30 '16 at 9:19