| Abstract: |
| Assessment of models should depend on the modeling objectives. Models that incorporate more
realistic mechanisms are more suitable for providing insights. To produce reliable and accurate
predictions and inform public health decision making, parsimonious models are more appropriate
and the modeling needs to respect the data. Most of all, model calibration results should be validated
by data that is independent of the calibration data, before scenario analysis is made to inform policy.
As a case study, we revisit the well-known example of a 1978 influenza outbreak in a boarding school in England.
We demonstrate that a parsimonious SIR model with data informed time-dependent parameters can
produce both accurate fitting to the time series data and validation by the final size of the epidemic.
Furthermore, modeling results also provide evidence of the likely epidemic control measures implemented
at the school. |
|