njtierney
commented
1 month ago For example, there are existing surveys listed in socialmixr, which get the social contact surveys that can then be used - see https://zenodo.org/communities/social_contact_data?q=&l=list&p=1&s=10&sort=newest
Some other TODOs on this topic:
- [x] Use another survey from somewhere in Europe (a comparison point for POLYMOD)
- [x] Another survey country from a country that has a different contact pattern to POLYMOD, e.g., Zambia, Indonesia, Singapore.
- [x] Fit models using POLYMOD, POLYMOD-alike, and other country.
- [ ] Use those fitted models to then predict (create a contact matrix) to a country like Australia, and see how the contact matrices are different across different contact surveys. So we are keeping the "newdata" part of the prediction the same, and changing the models that are used.
Some notes about communication of this
- We step between model fitting and model prediction a lot in this, so it will be important to make it as clear as we can how those processes are different
- Potentially pre-compute the different models from the different surveys.
- Discuss some rules/heuristics on how to choose a contact survey that will might be best for your country/population
- It will be important to discuss and explain to people why using models that are build on contact surveys that are quite different to their country (different meaning different contact/social patterns within a country).
- For example, using a model fit using the POLYMOD contact survey, which is based in Europe, on a multi generational population like Singapore, will probably give you quite different results to using a contact survey from Singapore. (see the paper,"Apparent structural changes in contact patterns during COVID-19 were driven by survey design and long-term demographic trends" for a discussion on this problem: https://arxiv.org/abs/2406.01639)
Add to the vignette "Using other data sources"
Potentially add a new class of data that helps define a contact survey.