Serrien, Ben
keywords: soil; EU; soil monitoring law; transfer functions; pedotransfer functions; statistics
Deliverable D5.1 – Statistical analysis plan for best practices for the development and validation of TF and PTF.
From the grant agreement, Annex 1, part B:
As a first step, WP5 will develop a statistical analysis plan based on input from WP1, i.e., determining statistical practices in use regarding TF and PTF (supply-side analysis) and the requirements of the call and the SML (demand-side analysis). For the supply-side analysis, partners will determine which information needs to be extracted from published sources related to TF/PTF to make a statistical assessment of their quality and applicability. This will include, a.o., algorithm used, required input data, characteristics of the training data, uncertainty of the estimation, type of validation (non, cross-validation, or independent validation set). The various statistical methods will then be first divided into methods related to method comparison; methods associated with TF development and methods related to PTF development. Their underlying assumptions will be made explicit, and conceptually similar models/methods will be grouped. For instance, within TF models, a distinction will be made between asymmetrical (e.g. linear regression models) and symmetrical models (e.g. orthogonal or major axis regression, Deming regression, ...). Relevant literature on similar statistical techniques will also be drawn from other scientific domains. For the demand-side analysis, partners will translate the requirements of the call and SML to clearly defined research questions to which a specific statistical task can be attributed. The demand and supply sides will then be confronted with each other and synthesized, resulting in a list of methods suitable for answering the various research questions and statistical tasks. Potential knowledge gaps regarding statistical methods will be identified. Each of the research questions will naturally depend on the others, reflecting the TF and PTF life cycle (development, validation, use), which will form the basis of the statistical analysis plan. Best statistical practices will be identified and applied in the statistical analysis plan, which forms the backbone of WP5 tasks.
This github repo will be used for writing the statistical analysis plan.
The project Towards a harmonised pan-European monitoring of soil health descriptors, also known as SOILHARMONY, receives funding from the European Union’s HORIZON Innovation Actions 2022 under grant agreement No. 101296615.