Metamodeling

Metamodeling involves a suite of analytical tools, including artificial intelligence in conjunction with simulation modeling.

Metamodel = model of a model!

Metamodeling involves using analytical tools, such as regression modeling or artificial neural networks on mechanistic simulation models. These tools are capable of advancing the science of modeling in public health via increasing computational efficiency of simulation models or revealing the characteristics of simulation models.

For an example of a metamodel in practice, check the project ModEx which uses a probabilistic sensitivity analysis dataset to reveal model characteristics.

References

2021

  1. baycann.png
    BayCANN: streamlining Bayesian calibration with artificial neural network metamodeling
    Hawre Jalal, Thomas A Trikalinos, and Fernando Alarid-Escudero
    Frontiers in Physiology, 2021

2015

  1. Computing expected value of partial sample information from probabilistic sensitivity analysis using linear regression metamodeling
    Hawre Jalal, Jeremy D Goldhaber-Fiebert, and Karen M Kuntz
    Medical Decision Making, 2015

2013

  1. metamodel.png
    Linear regression metamodeling as a tool to summarize and present simulation model results
    Hawre Jalal, Bryan Dowd, François Sainfort, and Karen M Kuntz
    Medical Decision Making, 2013