Project collaborators
Damon Smith, Department of Plant Pathology
Maria Oros, DSI
Project start and end dates
7/1/23-11/30/23
Project summary
This project addressed the rising use of foliar fungicides on corn crops, investigating whether these fungicides were effectively controlling diseases and potentially enhancing overall plant health. From 2019-2023, university researchers across 18 U.S. states applied fungicides at different growth stages of corn to determine their impact on yield and disease control and used a rigorous statistical analysis to assess their effectiveness. This approach allowed the researchers to provide concrete, evidence-based insights into how and when fungicides should be used for optimal results.
The study was designed using a uniform protocol with a randomized block structure, where each treatment varied from 4 to 6 repetitions. To evaluate the efficacy and subsequent yield effects of various common fungicides and their application timings, we utilized a Network Meta-Analysis (NMA) approach, comparing each treatment to a non-treated control (NTC). We modeled the yield effect size as the objective variable, incorporating the inverse of the within-study variance as the weight for each study. Fungicide application and timing were treated as fixed effects, while the variability between studies was addressed by considering the study as a random effect due to significant between-study variability expected from diverse methods and locations. A heterogeneous variance-covariance structure based on the heterogeneous compound symmetry model (CSH) was utilized.
We employed the standard type III test for fixed effects to determine the significance of fungicide application and timing at the 5% level. Least Squares Means were calculated to estimate the mean effects across all studies, along with their standard errors and 95% confidence intervals. A standard normal test assessed the statistical significance of these mean effects. Pairwise comparisons were evaluated using Fisher’s least significant difference method, with results presented through a pairwise letter labeling system.
The analysis was initially conducted in SAS, followed by further validations in R using the lme4 package and netmeta for comprehensive meta-analytical assessments. The model’s performance was evaluated using the Akaike Information Criterion (AIC) and the Chi-squared test per degree of freedom. Additionally, post hoc power tests were conducted to confirm statistical significance and resolve any discrepancies observed in the pairwise comparisons.
The impact of this research is significant: it confirms the effectiveness of foliar fungicides in managing corn diseases and improving yields. The primary beneficiaries of this work are researchers, agronomists, farmers, and policymakers involved in corn production. These stakeholders gain valuable information that can influence agricultural practices and decision-making processes related to crop management. The study’s findings have been disseminated through academic publications, conferences, and outreach activities, ensuring that the agricultural community is well-informed. This knowledge empowers stakeholders to adopt better crop management strategies, directly enhancing corn production outcomes.
Project deliverables
- SAS macro framework
Our SAS macro framework is designed to automatically calculate and report critical statistical metrics such as mean estimates, confidence intervals (CIs), and p-values, and perform pairwise comparisons and post-hoc power tests. By centralizing these functions within a single, comprehensive macro, researchers can efficiently process data while maintaining high analytical standards. This setup not only streamlines the analytical process but also ensures that the statistical outputs are directly aligned with the scientific requirements of meta-analytical studies in agriculture. - Streamlit dashboard for visualization and meta-analysis
The modeling framework is complemented by a sophisticated Streamlit dashboard, which serves as the primary interface for visualizing model results and key statistics. - Scientific paper
Work is ongoing on a scientific paper that details the methodology and insights derived from using these advanced analytical tools.