Urban Oxidative Stress Avian Meta-Analysis Data
Description
These data include 137 effect sizes from 26 different studies. Each effect size represents a biomarker of oxidative stress. Most are paired means (one urban site and one non-urban site), while a smaller subset (3 studies, 25 data points) are correlation coefficients representing an urban gradient.
Files
Steps to reproduce
Literature Search and Data Collection--- We collected articles through systematic review and literature searches using a combination of the keywords avian, oxidat*, stress, bird, urban. We searched the databases “Web of Science” and “Google Scholar” in July 2020, and “Scopus” in May 2023. Through this search and literature gathering process, we identified 175 articles. To be eligible for consideration, an article had to meet the following criteria: 1) study on birds of wild origin, not a captive breeding population, 2) compare urban and non-urban populations of the same species in the same study, 3) report some measure of oxidative stress, including oxidative damage, antioxidant capacity, or levels of specific measured antioxidants. Thirty-four articles met these criteria. We extracted data from articles directly from tables, supplemental material, and text-reported values where possible. If this was not possible, we estimated means and standard errors from plots using WebPlotDigitizer 4.6 [https://apps.automeris.io/wpd/]. If no usable plots were available, we contacted the authors of articles to request access to their data. We obtained usable data from 26 total articles. We also recorded the following variables from each article to be used as moderators in modeling analysis: species, age, tissue (from which oxidative stress was evaluated), sex, season, and absolute latitude (averaged between study sites). Using the R Script to Calculate Effect Sizes--- Use the R script provided, named MetaAnalysis.R, to calculate Hedge’s g. This script uses the package esc to calculate Hedge’s g and associated standard error for each data point. Most studies reported means and standard deviations from two paired populations. In this case, the script calculates Cohen’s d and then converts it into Hedge’s g. Three studies (total of 25 data points) reported correlation coefficient r across an urban gradient. For these studies, the script converts r to Hedge’s g: first we estimate a 95% confidence interval for the reported r based on sample size using the R package psychometric, convert r values into Cohen’s d and then Hedge’s g using the R package esc, and calculate the standard error of g based on the 95% confidence interval.