Omitted variable bias in studies of plant interactions

Area of Science:

• Ecology
• Plant Biology
• Environmental Science

Background:

• Models of plant-plant interactions are crucial for understanding species coexistence, invasive species, and climate change impacts.
• Recent studies show predictive failures in competition models, questioning prior research.
• Unmodeled environmental heterogeneity (e.g., nutrients, soil pathogens) is a likely cause of these model failures.

Purpose of the Study:

• To test for and correct omitted variable bias in plant competition studies.
• To assess the impact of unmodeled environmental heterogeneity on competition estimates.
• To improve the accuracy of plant interaction models.

Main Methods:

• Utilized instrumental variables analysis to address omitted variable bias.
• Applied methods to studies following common protocols for measuring plant competition.
• Conducted an observational study and a quasi-experiment with controlled competitor seeding.

Main Results:

• Omitted variables caused plant competition to appear as mutualism in an observational study.
• In a quasi-experiment, omitted variables underestimated competition by approximately 35%.
• Bias was observed even in a relatively homogeneous agricultural field setting.

Conclusions:

• Accurate estimation of plant competitive interactions remains challenging due to pervasive omitted variable bias.
• Environmental heterogeneity significantly distorts perceived plant competition.
• True experiments with complete control over competitor abundance are ideal but rare, highlighting the need for robust statistical methods like instrumental variables analysis.