In which of the following situations is it best to use the historical simulation method when estimating VAR?
1) When the data is normally distributed
2) When the data is not normally distributed
3) When the data is missing values
4) When the data is highly correlated
