Rectifying long-standing misconceptions about the ρ statistic for molecular dating

Vincent Macaulay, Pedro Soares, Martin Richards

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

When divided by a given mutation rate, the ρ (rho) statistic provides a simple estimator of the age of a clade within a phylogenetic tree by averaging the number of mutations from each sample in the clade to its root. However, a long-standing critique of the use of ρ in genetic dating has been quite often cited. Here we show that the critique is unfounded. We demonstrate by a formal mathematical argument and illustrate with a simulation study that ρ estimates are unbiased and also that ρ and maximum likelihood estimates do not differ in any systematic fashion. We also demonstrate that the claim that the associated confidence intervals commonly estimate the uncertainty inappropriately is flawed since it relies on a means of calculating standard errors that is not used by any other researchers, whereas an established expression for the standard error is largely unproblematic. We conclude that ρ dating, alongside approaches such as maximum likelihood (ML) and Bayesian inference, remains a useful tool for genetic dating
LanguageEnglish
Article numbere0212311
JournalPLoS One
Volume14
Issue number2
DOIs
Publication statusPublished - 19 Feb 2019

Fingerprint

Maximum likelihood
statistics
Statistics
Likelihood Functions
mutation
Mutation Rate
Uncertainty
confidence interval
uncertainty
researchers
Research Personnel
Confidence Intervals
Mutation
phylogeny
sampling

Cite this

@article{ad47bb3b827849dea53f89f2eb49bb7e,
title = "Rectifying long-standing misconceptions about the ρ statistic for molecular dating",
abstract = "When divided by a given mutation rate, the ρ (rho) statistic provides a simple estimator of the age of a clade within a phylogenetic tree by averaging the number of mutations from each sample in the clade to its root. However, a long-standing critique of the use of ρ in genetic dating has been quite often cited. Here we show that the critique is unfounded. We demonstrate by a formal mathematical argument and illustrate with a simulation study that ρ estimates are unbiased and also that ρ and maximum likelihood estimates do not differ in any systematic fashion. We also demonstrate that the claim that the associated confidence intervals commonly estimate the uncertainty inappropriately is flawed since it relies on a means of calculating standard errors that is not used by any other researchers, whereas an established expression for the standard error is largely unproblematic. We conclude that ρ dating, alongside approaches such as maximum likelihood (ML) and Bayesian inference, remains a useful tool for genetic dating",
author = "Vincent Macaulay and Pedro Soares and Martin Richards",
year = "2019",
month = "2",
day = "19",
doi = "10.1371/journal.pone.0212311",
language = "English",
volume = "14",
journal = "PLoS One",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "2",

}

Rectifying long-standing misconceptions about the ρ statistic for molecular dating. / Macaulay, Vincent; Soares, Pedro ; Richards, Martin.

In: PLoS One, Vol. 14, No. 2, e0212311, 19.02.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Rectifying long-standing misconceptions about the ρ statistic for molecular dating

AU - Macaulay, Vincent

AU - Soares, Pedro

AU - Richards, Martin

PY - 2019/2/19

Y1 - 2019/2/19

N2 - When divided by a given mutation rate, the ρ (rho) statistic provides a simple estimator of the age of a clade within a phylogenetic tree by averaging the number of mutations from each sample in the clade to its root. However, a long-standing critique of the use of ρ in genetic dating has been quite often cited. Here we show that the critique is unfounded. We demonstrate by a formal mathematical argument and illustrate with a simulation study that ρ estimates are unbiased and also that ρ and maximum likelihood estimates do not differ in any systematic fashion. We also demonstrate that the claim that the associated confidence intervals commonly estimate the uncertainty inappropriately is flawed since it relies on a means of calculating standard errors that is not used by any other researchers, whereas an established expression for the standard error is largely unproblematic. We conclude that ρ dating, alongside approaches such as maximum likelihood (ML) and Bayesian inference, remains a useful tool for genetic dating

AB - When divided by a given mutation rate, the ρ (rho) statistic provides a simple estimator of the age of a clade within a phylogenetic tree by averaging the number of mutations from each sample in the clade to its root. However, a long-standing critique of the use of ρ in genetic dating has been quite often cited. Here we show that the critique is unfounded. We demonstrate by a formal mathematical argument and illustrate with a simulation study that ρ estimates are unbiased and also that ρ and maximum likelihood estimates do not differ in any systematic fashion. We also demonstrate that the claim that the associated confidence intervals commonly estimate the uncertainty inappropriately is flawed since it relies on a means of calculating standard errors that is not used by any other researchers, whereas an established expression for the standard error is largely unproblematic. We conclude that ρ dating, alongside approaches such as maximum likelihood (ML) and Bayesian inference, remains a useful tool for genetic dating

UR - http://www.scopus.com/inward/record.url?scp=85061852660&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0212311

DO - 10.1371/journal.pone.0212311

M3 - Article

VL - 14

JO - PLoS One

T2 - PLoS One

JF - PLoS One

SN - 1932-6203

IS - 2

M1 - e0212311

ER -