Change over time & network comparisons more easy to make on map

[mlbtc]

We'd like to add ways to make changes over time & comparisons between different networks on the map easier to interpret. (For example: toggling between all jobs access vs. jobs by low-transit trips access, or pinning choropleth values to one standard map)

[mlbtc]

E.g. Access to jobs by low-fare trips only PINNED to access by all trips categories Travel on weekday evenings PINNED to travel at AM peak categories Access to low-wage jobs PINNED to access to all jobs Access in equity neighborhoods PINNED to access in all neighborhoods

[wklumpen]

Because of the additional data transfer and framework change that arises from trying to query/pull 2 data sets at once, we are suggesting the following middle-ground solution:

• Using the existing data for each region for each score, we calculate a fixed set of 5 bins that are appropriate for that measure (or measures, if there is a desire to combine jobs/low-income jobs or fare/non-fare into a single set of bins). This would be done using an external script.
• We store these fixed bins in a database table as a reference, and use them to display all measures with the appropriate colors across all region types.

This would also help to address #28 as it would allow for people to change geographies without the colors changing. It would also remove some computation time (since right now the percentiles are being calculated on the fly).

If there is still a desire to enable people to perform more direct comparisons using 2 (or more) datasets, I would suggest we put together a ~10-minute video on how you can download two data sets and make a quick comparison using qGIS. I would be happy to produce such a video if there is a desire for it.

[mlbtc]

I think this middle-ground approach will work well, thanks for suggesting. Would also be worthwhile to demonstrate how to make a comparison map as well.

[wklumpen]

I have calculated fixed bins for all regions and all scores based on the current data - we can easily update it with future data. To create "pretty breaks", we used the Jenks method instead of focusing on quantiles, which searches for clean breaks in the data and uses those to create bins.

This results in more geographical area falling into the worst category for cumulative measures (as the block groups have a long-tail distribution), but it allows for a visualization of how access has shrunk/grown in various situations. For example:

February 2020 in DC (total jobs in 60 min)

Versus July 2020 in DC

[wklumpen]

I created a separate issue for the video, which can be found at #134. Going to close this to mark complete, feel free to re-open if you have adjustments or suggestions.

[mlbtc]

This looks good! The ability to compare a few different scenarios on the fly because they have the same legend will be really useful and a good time saver on the site. Thanks!