We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ME

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Methodology

Title: Bayesian non-parametric inference for $Λ$-coalescents: consistency and a parametric method

Abstract: We investigate Bayesian non-parametric inference of the $\Lambda$-measure of $\Lambda$-coalescent processes with recurrent mutation, parametrised by probability measures on the unit interval. We give verifiable criteria on the prior for posterior consistency when observations form a time series, and prove that any non-trivial prior is inconsistent when all observations are contemporaneous. We then show that the likelihood given a data set of size $n \in \mathbb{N}$ is constant across $\Lambda$-measures whose leading $n - 2$ moments agree, and focus on inferring truncated sequences of moments. We provide a large class of functionals which can be extremised using finite computation given a credible region of posterior truncated moment sequences, and a pseudo-marginal Metropolis-Hastings algorithm for sampling the posterior. Finally, we compare the efficiency of the exact and noisy pseudo-marginal algorithms with and without delayed acceptance acceleration using a simulation study.
Comments: 28 pages, 3 figures
Subjects: Methodology (stat.ME); Probability (math.PR); Statistics Theory (math.ST); Populations and Evolution (q-bio.PE); Computation (stat.CO)
MSC classes: Primary: 62M05, Secondary: 62G05, 92D15
Journal reference: Bernoulli 24(3):2122-2153, 2018
DOI: 10.3150/16-BEJ923
Cite as: arXiv:1512.00982 [stat.ME]
  (or arXiv:1512.00982v5 [stat.ME] for this version)

Submission history

From: Jere Koskela [view email]
[v1] Thu, 3 Dec 2015 08:03:36 GMT (533kb,D)
[v2] Wed, 16 Dec 2015 15:32:37 GMT (533kb,D)
[v3] Wed, 13 Jul 2016 11:03:21 GMT (536kb,D)
[v4] Wed, 24 Aug 2016 16:19:58 GMT (532kb,D)
[v5] Mon, 23 Jan 2017 18:21:10 GMT (537kb,D)

Link back to: arXiv, form interface, contact.