ShiVa: Detection of Evolutionary Shifts in both Optimal Value and Variance

Version: 1.0.1
Authors:

• Wensha Zhang wn209685@dal.ca
• Lam Si Tung Ho Lam.Ho@dal.ca
• Toby Kenney tb432381@dal.ca

Overview

ShiVa is an R package for detecting evolutionary shifts in both optimal trait values (mean) and diffusion variance across phylogenetic trees, based on the Ornstein-Uhlenbeck (OU) model. While many existing methods detect shifts in optimal values only, ShiVa simultaneously identifies both types of shifts using a penalized likelihood framework with L1 regularization.

The method is designed for researchers in evolutionary biology and comparative methods, particularly those interested in identifying abrupt regime changes along a phylogeny.

See our preprint for methodological details: arXiv:2312.17480

Features

• Automatively estimates shifts in both mean (θ) and variance (σ²) under the OU model
• Supports model selection using BIC, mBIC, or pBIC
• Includes backward correction for improved parsimony
• Can optionally account for measurement error
• Fast optimization using coordinate descent and soft-thresholding

Installation

To install the package from GitHub:

# install.packages("devtools") # if not already installed
devtools::install_github("WenshaZ/ShiVa")