Confidence intervals for semi-parametric empirical Bayes
D. R. Bickel, “Large-scale interval and point estimates from an empirical Bayes extension of confidence posteriors,” Technical Report, Ottawa Institute of Systems Biology, arXiv:1012.6033 (2010). Full preprint
To address multiple comparison problems in high-dimensional biology, this paper introduces shrunken point estimates for feature prioritization and shrunken confidence intervals to indicate the uncertainty of the point estimates. The new point and interval estimates are applied to gene expression data and are found to be conservative by simulation, as expected from limiting cases. Unlike the parametric empirical Bayes estimates, the new estimates are compatible with the semi-parametric approach to local false discovery rate estimation that has been extensively developed and applied over the last decade. This is carried out by replacing strong parametric assumptions with the confidence posterior theory of papers in the presses of Biometrics and Communications in Statistics — Theory and Methods.
David Bickel, statistician 
The proposed approach has the advantage that its interval and point estimates are not biased toward a conditional prior given the alternative hypothesis. Such bias is undesirable since the prior often cannot be estimated accurately.
In addition, the proposed interval estimates coincide with the usual confidence intervals when the LFDR estimate is small.
Published version:
http://goo.gl/v7TWR