I work with a dataset that gathers stats from all across a city. Each datapoint is of interest and worth visualising to our users, but there are a few parts of the city that significantly deviate from the average into hostspot clusters.
For example many wards throughout the city would have a value of [0-5] for any given month but one particular ward might be [300-400+]. Additionally there's not much mid range.
I'm trying to think about clever ways of visualising this that demonstrate both the data and it's month-by-month changes for most wards throughout the city, but also clearly demonstrate how much of an outlier particular wards can be.
I've tried:
Heatmaps - Radius needs to be set to a narrow zone to prevent spill over small boundaries, but due to the massive deviation of our outliers I end up with very pale, widely distributed wards across the city and then one tight dark blotch in the middle. It looks like a pitri dish and isn't very effective. Increasing the radius improves readability but means that datapoints now spill over ward/borough lines.
Choropleth - Much more visually appealing but again in a single swatch of colours (yellow for our [0-5]'s and dark red for our [300+]) I think it's still difficult to demonstrate the magnitude of the skew. My mind draws a linear line between each colour step.
Really I need a good idea for geographically representing data on a Logarithmic scale when I have a massive deviation in my data. Additionally data for the mode is quite low significance (0-5) and so is subject to a high variability...
Really struggling to think about how best to plot this. Perhaps the answer not to try and instead break it into separate graphs demonstrating each aspect (skew, significance of mode and peaks?).
Has anyone tried mapping a dataset like this before?
