Causal Graph
Describe how to use the "Causal Graph" visualization in PopHR
The causal graph visualization is useful when trying to find the causal relationship of public health concepts and their associations estimated from our data.
Page layout
The page contains the control bars (top and bottom), the title of current visualization (top left), the legend for the graph (top right), and the Causal Graph centered on a selected concept (the main body of the page). The graph below shows the annotations of layout of the causal graph view and the annotations are directed by the blue arrows.
Features
The graph below shows the annotations of the features in the causal graph view and annotations are directed by the blue arrows.
What the plot represents?
The central concept of the Causal Graph is the selected concept highlighted in teal, for which causally-linked concepts are displayed. Above the node for the central concept are two buttons to either expand/collapse the network to display/hide other indirectly linked concepts.
Below the concept node is the label for the selected indicator and for each concept there is a default indicator pre-selected. A drop-down list can be accessed from the downward arrow if more indicators are available. To the right of the indicator label is a link ("i") to the methodology documentation for the selected indicator.
To the left of the central concept are concepts known to be upstream determinants or risk factors of the concept, while downstream consequences or burden are shown to the right.
How causal graph is built?
The Causal Graph (above) displays public health concepts that are causally linked to the central concept, based on consensus knowledge of public health experts.
How associations are calculated?
The associations between the default indicators of the linked concepts are calculated using Spearman's Rank-Order Correlation, which measure the strength and direction of association between two ranked variables.
In the causal graph view, two indicators are ranked among their highest shared geographic resolution available. For example, if both prevalence of diabetes and prevalence of hypertension are available at health region level, then the ranking is based on indicator values among all health regions. Users can change the indicator of a concept using the drop-down indicator list, and the association will be updated using the new indicator rank.
The color of the arrow describes the direction of associations, red for positive and blue for negative, while the thickness of the arrow describes the strength, bold for strong (ρ > ? ) and thin for weak (ρ < ? ). A grey dashed line describes null or non-significant associations. If associations cannot be calculated between two concepts or indicators due to a lack of shared geographic resolution, or a lack of data in general, the arrow between concepts remains solid black. This indicates that a known causal link exists bases on expert consensus, but cannot be validated with our data.
Last updated
