Date de l'évènement :
Webinar organisé conjointement par le SESSTIM et l'Open Health INSTITUTE.
The medical world is drowned by a tsunami of data, important in terms of size, but even more because of the complex relationships between them. It is therefore necessary to analyze these big data through a prism of methods, aiming to compress them cleaverly and to interpret them, using more or less structured models. These methods extract the links between the data in terms of:
- statistical correlation (like the classical data analysis)
- functional interaction (network analysis)
- temporo-spatial dependence (chronological and spatial analysis).
The compression retains only the data explaining the "weight" of the data (type factor analysis) and their most constant links in time and space.
Interpretation leads to temporo-spatial models of evolution that fit into the three currently available paradigms:
- the differential model, if the data are well resolved spatially and temporally
- the automata model, if the data can be reduced to "all or nothing"
- the multi-agent model, if the qualitative exchanges between actors creating the data are more important than the quantitative data flows.
Examples will be taken from the main sources of data, namely social data (example of obesity) and biomedical ones (example of genetic regulation and its pathological dysfunctions).
