Citation

Spjøtvoll E. Biometrika 1977; 64(2): 327-334.

Copyright

(Copyright © 1977, Biometrika Trust, Publisher Oxford University Press)

DOI

10.1093/biomet/64.2.327

PMID

unavailable

Abstract

A stepwise multiple comparison method similar to the Newman-Keuls method is proposed for the case where it is known a priori that the parameters are ordered. It is shown how the significance levels at the various steps can be chosen to keep the probability of at least one false rejection less than a given upper bound. The method can be used for both continuous and discrete distributions since it is essentially based upon tests ordering any two of the parameters.