{\rtf1\ansi\ansicpg1252\deff0\deflang1031{\fonttbl{\f0\fswiss\fcharset0 Arial;}} {\*\generator Msftedit 5.41.15.1507;}\viewkind4\uc1\pard\f0\fs20 Coverage intervals versus Likelihood ratio intervals\par G\'fcnter Zech, Universitaet Siegen, Germany, email: zech@physik.uni-siegen.de\par Abstract:\par An example is constructed such that a coverage interval and a likelihood interval disagree by a large extent. It is demonstarted that the likelihood interval is at least intuitively much more attractive. The reason for this behavior is the fact that coverage intervals accept all parameter values which are compatible with the measurement whereas likelihood ratio intervalls take into account the fact that only one and not several parameters can be true. As a consequence of this caviat of the coverage intervals which is related to a violation of the likelihood principle one should not require coverage for likelihood ratio or Bayesian intervals while coverage intervals which exclude relatively high likelihood ratios are very problematic. \par }