“When there are variables that are not normal in your dataset, we shall transform them to be normal” This is one of the most common points i see that has to be done before predictive modelling.
A. For instance if a dataset has got 2 variables that are not normal and if it is mixed data do we still transform the dataset,? That is if your dataset has got both continuous and categorical variables?
B. If the data has to be transformed, should this be done for all prediction models? Because, my understanding is that it is not required for models like logistic regression.
Please help me in understanding this concept. Thanks