2025/05/19 by Piepho, Hans-Peter, Claß-Mahler, Ingrid, Zimmernann, Beate +3
Mathematics · Environmental Science · Decision Sciences · #Advanced Causal Inference Techniques #Sustainable Agricultural Systems Analysis #Optimal Experimental Design Methods
paper · doi:10.5073/jfk.2025.01.20
The project “Sustainable agriculture 4.0 without chemical synthetical plant protection” (NOcsPS) tests a number of cropping systems that avoid the use of chemical synthetical pesticides while at the same time using mineral fertilizers. The experiment started in 2020 (sowing in fall 2019). In 2024 (sowing in fall 2023), some of the cropping systems were modified. Analysis of this experiment may be done using linear mixed models. In order to include the data from 2020 to 2023 in joint analyses with the data collected for the modified systems from 2024 onwards, the mixed modelling approach needs to be reconsidered. In this paper, we develop models for this purpose. A key feature is the use of network meta-analytic concepts that allow a combination of direct and indirect comparisons among systems from the different years. The approach is first illustrated using a toy example. This is followed by detailed analyses of data from the two trial sites Dahnsdorf and Hohenheim. Our proposed methodology can be applied in other long-term experiments as well and provides a general strategy for modelling changes of treatments. Our general recommendation is to be very conservative in the conduct of long-term experiments and make changes only in very well justified cases, because the subsequent analysis and interpretation of results is complicated, as demonstrated in our paper.