Main Page
A sandbox for modeling how we might map evidence to theory in ecology - ZiF working group 2024/2025.
Current graph
The current graph models the following papers:
- Sod translocation to restore habitats of the myrmecophilous butterfly Phengaris (Maculinea) teleius on former agricultural fields
- Drivers of understory species richness in reconstructed boreal ecosystems: a structural equation modeling analysis
- On the specification of structural equation models for ecological systems
- Forest structure, plants, arthropods, scale, or birds’ functional groups: What key factor are forest birds responding to?
Current graph diagram
Causal properties
This sandbox has two causal properties (for now):
- mayCause (P1): A (<subject>) <may cause> B (<object>)
- hasCausalRelationWith (P9): A has some causal relationship with B where the direction is unknown (either A causes B or B causes A)
Fuzzy Cognitive Map properties
There is a suggestion in Pearson and Clark-Wolf 2023 that the results of Structural Equation Models can be used to 1) reduce model uncertainty and 2) estimate interaction co-efficients. Therefore we have two new properties which must be qualified by extracting the standardized interaction co-efficient:
- interaction strength (P11): a number between -1 and 1 that represents the coefficient of interaction.
- increases(P12): If <subject> increases, <object> increases by (interaction strength)
- decreases (P13): If <subject> increases, <object> decreases by (interaction strength)
These properties can then be easily deployed in a program for Fuzzy Cognitive Mapping as suggested interaction weights and model structure items (e.g. as a toolbox of model components).
Conceptual properties
This sandbox has two conceptual properties to link ecological concepts to one another:
- subclass of (P7): this item is a subclass (subset) of that item; all instances of this item are instances of that item; different from P31 (instance of), e.g.: volcano is a subclass of mountain (and an instance of volcanic landform); K2 is an instance of mountain
- part of (P8): object of which the subject is a part
Attribution
Each causal link should be attributed to the study that proposed or established it. This is done by adding a qualifier to the causal property that uses the property citation (P5) to attach the link. The article must first be created as an item in this Wikibase, and linked to its item on Wikidata using Wikidata equivalent (P4).
Questions still to be addressed
- Is it possible to subset this graph for a specific context and do analysis using SEM or some other method to assess the direction of linkages?
- How do we encode assumptions into meta-models?
- How do we incorporate interventions (e.g. invasive species removal) into this?
- How do sub-classes work in the mathematics of causal models? (e.g. can you move relationships from sub-class to parent class?)
- Does "no statistically significant association" need to be documented?