data-integration models (dims)
DESCRIPTION
Data-Integration Models (DIMs). Bruce Edmonds, Centre for Policy Modelling, Manchester Metropolitan University. Underlying principle : Evidence should not be ignored without a very good reason. - PowerPoint PPT PresentationTRANSCRIPT
Data-Integration Models (DIMs)
Underlying principle : Evidence should not be ignored without a very good reason
Modelling Goal : to be consistent with a heterogeneous set of evidence at a level of detail which is as directly comparable to
that evidence as possible
In other words, to describe what is observed using a dynamic and detailed simulation model (rather than to predict or explain). Thus including kinds of data that are difficult to formalise, e.g. partial, qualitative, meso-level etc. This does not eliminate the need for independent validation, but Data-Integration Models can form a precise basis for subsequent abstraction stages or generalisation.
Data-Integration Simulation Model
Micro-Evidence Macro-Data
Abstract Simulation Model 1
Abstract Simulation Model 2
SNA Model Analytic Model
Data-Integration Models will be at the core of the SCID Project (Social Complexity of Immigration and Diversity) which will use these models as the start of an abstraction, analysis and evaluation process, forming a chain of
related models from descriptive to abstract. See http://www.scid-project.org
Bruce Edmonds, Centre for Policy Modelling, Manchester Metropolitan University
Advantages• To separate the processes of representation (evidence to DIM) and abstraction & generalisation (DIM to simpler model) • To have a consistent and precise integration of the evidence, maintaining the reference of details• To be a precise reference point for associated models and theories, experiments can be performed on DIM to check more abstract models (simpler simulation, statistical and analytic models)• Can be used to integrate very different kinds and levels of evidence/data e.g. narrative accounts, aggregate data, time-series data etc.
Disadvantages• Tend to be slow and complicated• Difficult to understand completely, need further analysis
Progressive abstraction from
the precise starting point of a DIM, to
other kinds of models (themselves models of the DIM)
Representing a variety of evidence in the DIM (micro
in the agent behaviour, macro
against the results)
An example of a DIM in use, as part of a chain of models
from evidence to
abstract theories
An example of a DIM in use, as part of a chain of models
from evidence to
abstract theories