I have a heavily positive distribution in my dataset of my x variable, the aim to put the variable into 4 categories of categorical variable. Although, my initial idea is to log the continuous data then split it into categories. However, lots of values of the dependent variable are 0 which then would require a constant in the log transformation. Is it appropriate to just split a positive skewed dataset into a categorical variable rather than transforming the data first.
Regards, Otis