https://github.com/gavinsimpson/ncse-seminar-2022

National Centre for Statistical Ecology seminar, Feb 9th, 2022
https://github.com/gavinsimpson/ncse-seminar-2022

biodiversity generalized-additive-models presentation rmarkdown splines trends

Last synced: 23 days ago
JSON representation

National Centre for Statistical Ecology seminar, Feb 9th, 2022

• Host: GitHub
• URL: https://github.com/gavinsimpson/ncse-seminar-2022
• Owner: gavinsimpson
• License: cc-by-4.0
• Created: 2022-02-09T08:25:59.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2022-02-10T09:19:32.000Z (over 2 years ago)
• Last Synced: 2024-06-11T19:56:58.848Z (5 months ago)
• Topics: biodiversity, generalized-additive-models, presentation, rmarkdown, splines, trends
• Language: HTML
• Homepage: https://gavinsimpson.github.io/ncse-seminar-2022/
• Size: 13.5 MB
• Stars: 3
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE
  • Citation: CITATION.cff

Awesome Lists containing this project

README

        
# NCSE Seminar 2022

## Quantifying trends in biodiversity with generalized additive models

### Gavin Simpson

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.6033546.svg)](https://doi.org/10.5281/zenodo.6033546)

Slides & related R code for a talk that gave as part of the National Centre for

Statistical Ecology seminar series, Feb 9th, 2022.

Rendered slidedeck: https://gavinsimpson.github.io/ncse-seminar-2022/

#### Abstract

Climate change and other human-caused environmental disturbance may lead to

declines in biodiversity. Recently, a number of studies have collated large data

sets of monitoring time series for selected ecosystem or organism groups and

used these data sets to estimate trends in biodiversity, with many studies

identifying large declines in biodiversity across a number of organisms or

ecosystems. These results are not without controversy however; data selection

and quality issues, as well as questions over statistical methodology have lead

to vigorous debate at meetings and in scientific journals.

Typically, trends in biodiversity are estimated using linear effects, via

generalized linear mixed (or hierarchical) models to account for site-to-site

heterogeneity in temporal trends. Additionally, year-to-year variation may

enhance or mask estimated losses or gains in biodiversity over time if the

first observation year in a given series is unusually rich or depauperate.

Using year random effects has been suggested as a mechanism to account for this

potential bias. An alternative – but related – way to model trends in

biodiversity time series is using penalized splines for the trends, leading to

hierarchical generalized additive models (HGAMs; also called structural additive

models). In this talk I’ll introduce HGAMs and penalized splines and their use

for modelling biodiversity trends, and illustrate the approach using an

arthropod time series data set.