1, Scenario:
(1) We are going to run a linear panel data regression；(2) some continuous variables of our model have a high value of skewedness (say, >3) and kurtosis (say, >3).

2, Questions
#1: should we do variable transformation to meet the requirements of variable normality? Some posters suggest that we do normality test after regression.
#2: what is the respective bench mark for skewedness and kurtosis? [-3, +3] for both? Besides, when should we transform a variable: when either its skewedness or kurtosis bypasses the benchmark or both do.

Many thanks!