Ten tips for using path analysis in IPM and biocontrol research

Sunday, November 13, 2022

11:12 AM – 11:24 AM PT

Location: Vancouver Convention Centre, Meeting Room 301

Presenting Author(s)

Robert E. Clark

Consulting Scientist
EcoData Technology
Oxford, Connecticut

Co-Author(s)

ER

Emily Rampone

Washington State University
Pullman, Washington
DC

David Crowder

Associate Professor
Washington State University
Pullman, Washington

Research on natural enemies & beneficial insects requires evaluating both the direct impacts of treatments on pests and the indirect effects on crops. Path analysis (structural equation modeling) is a powerful statistical tool for analyzing these data because it can model both direct and indirect interactions. Path analysis has been applied in many areas of ecological research and is growing in popularity in entomology and agroecology. Even though users have access to several new R packages, it can be challenging to implement this technique since it has some strict statistical assumptions. The goal of this talk is to show use-cases of path analysis with several examples from IPM and biocontrol research. We will then outline some important strengths and weaknesses of this approach that we have experienced in our own work. For example, path analysis requires that you know the direction of interactions (e.g. top-down or bottom-up) and this usually requires an understanding of all the feeding interactions in the system ahead of time. Highly complex field treatments with many distinct levels are difficult to model (e.g. categorical data with >3 levels). Additionally, including too many environmental variables can make path diagrams too convoluted and difficult to interpret. Often the best approach is to start with an "a priori" interaction network, and we will provide examples of how to do this. Overall, most restrictions can be addressed, but sometimes the best approach may be to use other multivariate techniques.