Group 1: Is the measure of Y axis of the graph the count of crashes? Did you consider to use something else (for instance number of fatalities)?
A: the number of fatalities is visible for a given crash into the tooltip (clicking on a marker). We found the number of crashes to be a good comparison indicator.
Group 3: What would your data look like if zoomed in? Can you access information for a certain accident?
A: When zoomed in, the cluster will gradually turn into individual crashes, which are represented by circle by default and can be changed to different markers as user's wish (weather, sex and light condition). And yes, you can access information for a certain accident by touching the marker on the map.
Group 4: How does multiple states reflect on the map as compared to graph & scatter plot?
A: in the map can be visualized one state at time, the last selected state is shown on the map. The markers are updated to match the last filter applied.
Group 5: How much g$%^&# did you put in for age? How was the y-axis of age mapped?
A: y axis in the graph always represent the count of the values matching a specific filter for a given year. In the filter age is clustered in decades.
Group 6: How is the clustering done on map and what amount does the shade represent?
A: clustering is performed based on a distance-based algorithm described in this page. For each marker, it looks at each cluster to see how far it is from the center of the cluster. If the distance is less than a maximum distance we specified and the cluster is the closest, then that marker is aded to the cluster. If the marker fails to be added to any cluster then a new cluster is created containing that marker. The shade of clusters represent their cardinality as explained in question 9.
Group 7: Why does the map display aggregate AND individual data when map mode changes? Was that a bug or intentional?
A: No matter whether the map mode changes, we display clusters and individual crash together in the map in many levels of scales. We think that if there are only some of crashes in a cluster, we should display individual crashes instead. That's why when in many cases you can see both clusters and crashes. In different level of detail, we have different threshold for 'cluster-to-crash' transition and when user zooming into enough level of detail, clusters will be totally replaced by individual crashes.
Group 8: Is there another way to select a state, as it seemed difficult to select a state with less values. Also, I will have to spend a couple of seconds to look around for the state. Can the state be selected from the map?
A: no, the tiled map is the only way to select a state. In the map, the relative position of states is maintained, and since the application is supposed to be targeted to an audience familiar with US geography we didn't thought it could have been an issue.
Group 9: How is the color of cluster determined?
A: They are determined by the number of crashes in it. The more crashes in that cluster, the more reddish it is.
Group 10: Why didn't you choose something other than a slider for the to/from year ranges? (Sliders are bad on the wall)
A: we found the slider to be the most intuitive way, it has to be noticed that we are not using controlP5 but a custom component that behaves way better.