Skip to content

Delphi CAG JSON format

Adarsh Pyarelal edited this page Aug 17, 2018 · 2 revisions

The JSON file has the following fields:

  • name: The name of the model
  • dateCreated: Model creation date.
  • variables: List of CAG variables/nodes. Each variable has the following properties:
    • name: Name of the variable
    • units: Units of the variable (just a placeholder for now)
    • dtype: Datatype (all variables are real-valued right now, boolean/categorical to come later)
    • arguments: CAG nodes that have an edge pointing from them to the variable.
    • indicators: List of indicators for the abstract concepts represented by the variables. Each indicator has the following properties:
      • name: The name of the indicator
      • dataset: Deprecated field, to be removed later.
      • source: Source of the indicator (FAO/WDI or CYCLES)
      • unit: Units of the indicator
      • mean: Mean value of the indicator
      • stdev: Standard deviation of the indicator
      • time: Time (caveat: accurate only up to a yearly resolution for now) that the mean value corresponds to.
  • timestep: Timestep of the model (set to 1.0 by default for now)
  • edge_data: List of edges with their data. Each edge has the following properties:
    • source: The source of the edge.
    • target: The target of the edge.
    • CPT: The conditional probability table corresponding to a histogram of the values of $\theta$ (See this writeup for more details.). The CPT has two fields:
      • theta: List of left bin edges for the theta histogram.
      • P(theta): The height of the bin, proportional to the value of the probability distribution function of theta at this point.
    • polyfit: A least-squares polynomial fit to the probability distribution function of theta (requested by SRI for use with the PRAiSE-WM systems). polyfit has the following properties:
      • degree: The degree of the polynomial that is fit (default = 7)
      • coefficients: A list of real-valued coefficients of the polynomial fit, ordered from low to high.
    • InfluenceStatements: A list of INDRA statements that correspond to the edge, serialized in JSON form (see here for more details. The InfluenceStatements contain the provenance information.
Clone this wiki locally