Which statistical test?

If you had a data set which exhibited both non-normally distributed and normally distributed data, which statistical test would you use and why?