Introduction

  • the validity of the statistical inference depends on the validity of the underlying assumptions
  • in particular: classical test procedures (\(t\)-tests, \(F\)-tests) depend on the underlying normality of the data
  • sometimes (for large samples), the central limit theorem may help
  • if not, use nonparametric tests
    • they do not depend on particular distribution of the data
    • price paid: less powerful (larger probabilities of making errors of the second kind)