Network visualization to discern patterns of relationships between symptoms in dementia

摘要: The multidimensional characterization of complex biomedical systems usually demands a large number cases in order to obtain reliable inferences. Even so, the participants many studies is relatively small as, for example, typical clinical trials. Here we suggest an approach based on network visualization, combined with resampling, discern patterns relationships among variables. We illustrate how this can be applied analyze changes multiple outcomes people dementia. between several dozens variables were represented by connectivity graphs, drawn calculating relative risk observing pair symptoms individual their co-occurrence chance only. statistical significance was calculated generating bootstrap sample. If null hypothesis (e.g., risks = 1 or equivalently, pointwise mutual information 0) rejected, vertices graph representing connected edge. edges (the degree connectivity) reflects stage cognitive impairment, worse dementia indicated lower connectivity. Arranging consistently allows characteristic profiles displayed; turn allow treatment effects discerned, at-a-glance pattern recognition.