Main methodological issues common to the three agronomic questions addressed with system experiments.
|Should we consider that the cropping system is applied either:
- on a plot with a different sequence term (temporal dimension)
- on all the plots that each receive a sequence term (temporal and spatial dimensions)
|If there is only one plot per year, the studied system has only a temporal component.
If the system has several crops present each year, the choice will be made according to whether or not hypotheses consider the interaction between crops and/or the overall contribution of all crops to the considered indicators of performance.
|Manipulation of data so that rows in the data table correspond to the statistical units associated with agronomic questions: the plot or all the plots of the cropping system.
If the format of the data frame has already been considered during the planning phase, and that the database easily allows extractions, this operation can be performed with ease.
|Should we use data collected at the plot level or at the block level?
|It depends on the question. To optimize the relationship between variables/indicators, the scale must be as close as possible to the underlying hypotheses of the statistical analysis.
|Tables 5, 6 and 7 show that to answer the three agronomic questions, we used the two levels.
|Can data from the different experimental sites in the network be gathered, and a cross-cutting analysis be conducted?
|Yes, if there is a common protocol and measurements/observations.
It is important not to give identical names in the platforms for elements that are in fact different. For example, a block can only be found in one platform.
|The platforms become modalities of a "platform" variable in a global data table.
Give individual names to the blocks and plots to make the models fit the experimental design
|Are there enough replicates?
|This question should be addressed during the planning phase
Power tests are very difficult to implement in system experiments, given the amount of data collected and variables/indicators on which tests can be applied. Practical constraints rarely allow for acceptable number of replicates, but it is advisable to get as many as possible in order to avoid attributing to cropping system performance that are in fact random.
The number of replications is evaluated by taking into account the spatial and temporal units that can be processed as blocks.
|Identification of all the components of the experimental layout: experimental sites, blocks, plots. This permits the identification of the smallest statistical unit (here, the plot) and to deduce the number of units under the same experimental conditions and the total number of units involved in a given statistical model.
|Can we answer agronomic questions that can arise after the planning phase?
|This aspect has to be taken into account when planning the experiment. This can be achieved by ensuring that the experiment has a control and sufficient replicates, and a database system sufficiently flexible to allow any new variable to be stored. Thus, data already present could be used to answer new questions. This question is crucial for long-term experiments.
|For each new question, consider whether the data collected are appropriate to answer it. This was the case for the question on the benefits of diversification with oilseeds and protein crops.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.