Bickel, David R.2016-09-212016-09-212016-09-19http://www.davidbickel.comhttp://hdl.handle.net/10393/35185https://doi.org/10.20381/ruor-143An invariant distribution of loss functions leads to reporting weighted geometric means of bounds of either odds or Bayes factors such as likelihood ratios as measures of evidence or to reporting a minimax-distance prior or posterior distribution.encombining distributionsmeasure of evidencestrength of statistical evidenceimprecise probabilityWorking Paper