Source author record

Erik van Zwet

Erik van Zwet appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Methodology

Catalog footprint

What is connected

2works

1topics

1close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2020arXiv

The Significance Filter, the Winner's Curse and the Need to Shrink

The "significance filter" refers to focusing exclusively on statistically significant results. Since frequentist properties such as unbiasedness and coverage are valid only before the data have been observed, there are no guarantees if we condition on significance. In fact, the significance filter leads to overestimation of the magnitude of the parameter, which has been called the "winner's curse". It can also lead to undercoverage of the confidence interval. Moreover, these problems become more severe if the power is low. While these issues clearly deserve our attention, they have been studied only informally and mathematical results are lacking. Here we study them from the frequentist and the Bayesian perspective. We prove that the relative bias of the magnitude is a decreasing function of the power and that the usual confidence interval undercovers when the power is less than 50%. We conclude that failure to apply the appropriate amount of shrinkage can lead to misleading inferences.

preprint2014arXiv

Partial exchangeability of the prior via shuffling

In inference problems involving a multi-dimensional parameter $θ$, it is often natural to consider decision rules that have a risk which is invariant under some group $G$ of permutations of $θ$. We show that this implies that the Bayes risk of the rule is {\em as if} the prior distribution of the parameter is partially exchangeable with respect to $G$. We provide a symmetrization technique for incorporating partial exchangeability of $θ$ into a statistical model, without assuming any other prior information. We refer to this technique as {\em shuffling}. Shuffling can be viewed as an instance of empirical Bayes, where we estimate the (unordered) multiset of parameter values $\{θ_1,θ_2,\dots,θ_p\}$ while using a uniform prior on $G$ for their ordering. Estimation of the multiset is a missing data problem which can be tackled with a stochastic EM algorithm. We show that in the special case of estimating the mean-value parameter in a regular exponential family model, shuffling leads to an estimator that is a weighted average of permuted versions of the usual maximum likelihood estimator. This is a novel form of shrinkage.